Novell Releases PostgreSQL for NetWare

An anonymous reader writes "Ever since Oracle announced they wouldn't port 9i to NetWare, Novell has been scrambling to find an enterprise-capable DB. Now it looks like they're settling on PostgreSQL. This follows their decision to ship Apache as the default web server for NetWare 6. Linux aficionados might sneer at an old workhorse like NetWare, but it's got more than 80 million client licenses worldwide, and it ain't going anywhere anytime soon."

Comments

Re:what is netware?

[iggymanz]

Hahah! you must be a young-un'! Novell Netware was THE workgroup network file/print server for the late 80's. The version 3.x of it had a stability and ease of administration that puts most Unix systems to shame (then Novell ruined it in the 4.x versions with unstable add-ons to do interoperation with other platforms)

They had market share because they could do what Microsoft could not at that time - make a server OS.

Re:This surprises me

[StArSkY]

Thanks, I will do that.

Re:What the heck?!

[Cato]

PostgreSQL is an open source project, not a product. The developers can port it to Windows in any way they want - it does run on Cygwin, which I use a lot for other tools. Why is it such a problem to use Cygwin? It's just a DLL, you wouldn't need the whole Cygwin environment on a production server, only on development machines.

You have no right to be angry at an open source project that is done by volunteers, usually in their spare time - if you really want a native Windows port, you can either help port it yourself or pay someone to do it.

Novel needs to do this ..

[oh]

I'm posing from a Novel site right now. Everyone here seems to be happy with netware for the most part. It works well with the corperate desktop (yes its windows), and like all OS' when its well maintained is pretty stable. The NDS tree had all the functionality that this site needs long before Microsoft's Active Directory was released.

The only reason anyone talks about moving away from Netware is application support. This porting of Open Source apps is a good thing for Novel. If they can ship enough applications, then people won't migrate away from Netware, and if they can increase market share then more people will develop on their platform.

This could also be a good thing for Open Source. With a new group of profesional developers working on the code they could make progress on those features that the Open Source product may be lacking. They will fix bugs.

If they are smart, they will keep the most of the code base the same. If they fork too far then they won't be able to include developments made from the community. Of course, that means fixes and features added by the Novel developers would be covered by the GPL and would be given back to the community.

This sounds like a good thing for both parties. Novel gets more software to run on their servers, making their servers more attractive to customers, selling more.

Open source gets any fixed and modifications that they make. Isn't this what open source (or free software) is about, you get access to the code for free to use any way you like, provided you give everyone access to the improvements you make.

Re:Novel needs to do this ..

[Penguin Follower]

PostgreSQL is under BSD license... so Novell doesn't have to contribute back... but it would be nice if they would.

Re:Novel needs to do this ..

[SN74S181]

Unless they want to end up with an orphaned fork of PostgreSQL that they alone can maintain, they will contribute back.

It isn't that alien a concept. It's how things often go. The FUD from GNU advocates often isn't relevant.

Re:Novel needs to do this ..

[Sturm]

Hehe.
From the above post:

"I'm posing from a Novel site right now."

I just knew all you Novel people were posers

Re:This surprises me

[C_Kode]

I don't consider PostgreSQL as an alternative to Oracle. PostgreSQL is a nice DB, but it isn't the be all end of of dbs. But then again neither is Oracle. If I had to choose between the two it wouldn't be cut and dry until you gave me the circumstances. Personally I use PostgreSQL in my business environment and it has worked flawlessly so far. Granted it doesn't run MC applications (Mission Critical) But it is used to make major business decisions (more or less data-warehousing)

I wouldn't dismiss PostgreSQL so quickly. Then again I wouldn't risk MC applications on it without further educating myself on it also. But it's that true for all major business decisions? I DBA 3 psql dbs. I'm happy with them. I even have live backup software for them. My DBA skills arn't great, but well enough to have judged correctly on what I have.

Any other PostgreSQL DBAs have farther exp on this subject?

Re:what is netware?

[hpavc]

i would have to say that novell is pretty rock solid ... once you get it up and running it has an amazing uptime. however when you mess with it (ala make a weekly arcserve upgrade or something) thats when it tends to be flaky.

it also doesnt like running out of diskspace or anywhere near out of diskspace.

with good hardware its very nice for a workgroup situation ... not that unlike samba :)

Re:What the heck?!

[realkiwi]

Get information before posting - it does runn on Windows.

In two flavors: with cygwin and a new native port that has branched recently

Slashdot is for facts not badly informed trolls

Re:This surprises me

[joib]

Actually, since version 7.1 (current is 7.2) row size is unlimited. Or, rather the limits are imposed by the operating system (2GB files on ext2?).

SQL Anywhere for Netware

[jfpoole]

If you're looking for a nice RDBMS for Netware, iAnywhere Solutions has SQL Anywhere, which is available for a number of platforms including Netware. I'm not exactly sure how it stacks up against PostgreSQL, but I've had a lot of success using it in the past (on Linux and Windows, admittedly).

Not that I work for iAnywhere Solutions or anything. *cough*

-j

Re:SQL Anywhere for Netware

[hey!]

For people who've been around long enough to remember Watcom SQL, this is the same product. It was acquired by Powerbuilder so they would have a product which played a similar role to PB that Interbase does for Delphi. Later when Sybase acquired Powerbuilder, they inherited this product, and decided to target it at the mobile market and workgroup application developers. Characteristically for Sybase been vascillating as to what to call this product: SqlAnywhere or Adaptive Server Anywhere.

SQL Anywhere is petty much in the same league as interbase. It has stored procedures, triggers, transactions and transaction logging, hot backup etc, and only cost about $1000 per ten users. It's quite stable (but stay away from .0.0 release). It has a very low footprint and performs fairly well. It's available on Linux Netware and Windows, and I believe we may see it on MacOS X soon.

Things I especially like about ASA: nearly meets ACID requirements (except concurrency, which it takes care of by record locking). It's relatively easy to support when you have non-experts managing the database -- good for application developers. There's a feature which allows you to store java objects in the database, which looks cool although it has no affect on me. It has excellent SQL*92 support. One really interesting thing is that since it has been acquired by Sybase, it has been getting Transact-SQL capabilities. This means you can easily port your database to Sybase Adaptive Server Enterprise or Microsoft SQL Server.

Things that I don't like about ASA: Behavior of the security system is a bit strange. I don't like the way it uses group membership to resolve unqualified table names -- you'd better create the table as the right user. The replication system is very complex and is not really an option for application developers because of support issues. This also means that for the mobile market, while the database is capable of runing on Palm Pilots and WinCE boxes, it's not in my opinion a very good choice for application developers.

Things I have mixed feelings about: having a relationship with Sybase. I like the SQL Anywhere people, but Sybase as a whole I neither like nor trust.

Finally, of course, SQL Anywhere is closed source, although Sybase has been reasonable with its licensing and pricing, and the product works reasonably well so it's not a huge issue if you have no political problem with proprietary software.

When we got started with our current application about six years ago, Sql Anywhere was a very good choice. Today, however, I'd seriously look at Postgres instead. SqlAnywhere's mobile deployment options could be a killer app for us, except don't think it is practical to deploy in relatively unsupported environments.

Re:SQL Anywhere for Netware

[hey!]

Our latest release (8.0.0) was fairly stable when it went out the door in December. A number of comments we got from our customers indicated that they were rather surprised at how stable it was (especially for a .0.0 release). Hopefully we'll do the same for the next major release as well.

Well, what constititues an 0.0 release is a bit arbitrary, but 7.0.0 was a clinker.

With respect to replication, I was basing this on 7.x. I have in fact managed to get replication working, but it was a butt-ugly process that I wouldn't want to put customers through. It seemed to me that there were lot of arbitrary constructs that you needed to master to get it working -- almost a science into itself.

This is why I've never used it for mobile app development, even though mobile apps are a big part of our business now. I want something I can deploy in a way that local admins don't have to go to sybase schoool to master. We're doing lots of work with West Nile virus surveillance, and we have a first class team to put mobile apps together.

If 8.0.x replication is easier for admins, I'd be very interested. Here's what I'd like to see: QT, Java or C#/Windows Forms on the PDA (in that order), with ASA for the database. This solution would absolutely ROCK, if it could be made push button simple for the local adminsitrators.

Otherwise for now we'll stick with Pumatech (although hosting the table in ASA tables would still be a good thing, if that were possible a la Oracle Lite).

MySQL vs. PostgreSQL

[mbogosian]

If you do a Google search for "MySQL vs. PostgreSQL, you'll get a lot of hits. Here are a few that seem to be pretty informative (if not slightly dated):

here
here
here
here
here
here (not really a comparison, but read this article and the linked Postgres article for more info)

In my personal experience, Postgres has historically been the database more prepared for larger, more multi-threaded applications.

Obviously, there have been debates about which are faster in various different applications. To be honest, I have no hard data, nor have I stretched them either to their capacity, but as a user and casual developer, they are both fast enough for me not to notice.

What's inarguable exciting can be directly quoted from MySQL's own comparison of the two (listed above):

[B]oth products are continually evolving. We at MySQL AB and the PostgreSQL developers are both working on making our respective databases as good as possible, so we are both a serious alternative to any commercial database.

Re:MySQL vs. PostgreSQL

[GooberToo]

MySQL supports a compressed client/server protocol which improves performance over slow links.

To the best of my knowledge, users have never requested that of PostgreSQL. Naturally, you can always tunnel a PostgreSQL connection through a compressed link using third-party products, but really slow links, such as modems, compress automatically anyway.

Actually, it has been asked for and the core developers rejected the notion. One of the commercial PostgreSQL companies currently have an working implementation (Mammoth PostgreSQL). I'm expecting it to be available commercially for 7.3 and then in open source fashion for 7.4.

There are many, many reasons why this is a good idea and many, many reasons why it's a bad idea to rely on modem compression.

For now, for those that need this feature, PostgreSQL is behind, however, hope is right around the corner. In the mean time, you'll have to rely on a kludgely solution by using SSH or some other tunnel that supports compression. There are, of course, several reasons why using a tunnel can result in less than ideal results.

Re:This surprises me

[Anonymous Coward]

> I thought Postgres would have too many limitations
> to be considered a healthy alternative to Oracle.
> eg 8k row sizes. Before people flame away. It has
> been a few years since I touched postgres, so this
> may be fixed by now.

Row sizes are now unlimited. See the Postgres
Limitations info page for more info.

I once felt the same way you did about Postgres, not enough features. I don't think it was until version 7.1 that postgres got outer joins (maybe only left/right outer joins, can't remember, but it was missing something like that). But recetly it has matured quite a bit and is a very nice database in my opinion. I use it for several applications, and I like it a lot.

There are still a couple of things that are a bit clumsy with PostGres (deleting a column, for instance), but I believe it has most major features you'd expect from an rdbms. At one of my jobs we use MSSQL7 for everything, and I've worked with it a lot. In my workings with recent versions of Postgres I have not stumbled across anything that I expected to be there from my experience with other DBs that was not in Postgres.

I'd give it another look if I were you. I trust Postgres a lot more now and I'm very pleased with it. It is working really well for me.

Ummm....

[Wee]

I've heard people swear on their souls that Postgre can stand up to MySQL, I look forward to finding out if this is really true. If it stands on it's own as a truly competitive product, congratulations to the Postgre guys. If however it's drawbacks are real, this could very well crush it forever.

Without sounding like I'm flamebaiting you, have you used many databases in your career? Do you know from experience the pros and cons of each? What drawbacks are you talking about? PostgreSQL is in a completely different class than MySQL. One is meant to be a full-fledged RDBMS, the other is meant to act as a super fast, network-aware DBM file on steroids. Each has their place, and they are more complementary than not. They can exist together, but you should never try to use one in place of the other. Get both, test both. Find the right tool for the job without listening to fanboy hype.

Oh, wait. IHBT. Never mind...

-B

Re:Ummm....

[Anthony Boyd]

I've heard people swear on their souls that Postgre can stand up to MySQL, I look forward to finding out if this is really true.

Without sounding like I'm flamebaiting you, have you used many databases in your career? Do you know from experience the pros and cons of each? What drawbacks are you talking about?

You didn't sound like you were flamebaiting him much, although I cringed as I read your post, waiting for the usual comment like, "MySQL is garbage even though everyone uses it, PostgreSQL is heaven and only the blessed use it." But you didn't quite do that, although calling MySQL a DBM file is a bit hostile.

So assuming we're being reasonable, here is what each side basically knows (and exaggerates) about the other. MySQL is supposedly feature poor, an awful thing without transactions, foreign keys, subselects, and other features you would find in Oracle. MySQL is for kiddies. Supposedly. Of course, most of the lacking features were implemented long ago or are about to see the light of day in MySQL 4.1. And most MySQL users freely admit they don't even need the features. MySQL gets deployed on fast-as-hell Web sites that only need to store data and display it. MySQL is for that Web site running on a Linux box that sees waaaayy too many SELECT statements during peak seconds. Cause that's what MySQL does best, and much to the disappointment of high-end database gurus, that's ALL most Web sites need. So the tool to do that best wins that market, and the PostgreSQL fans are just sour about that. On the other side, PostgreSQL is supposedly unstable and difficult. And PostgreSQL has some big-assed speed issues. Supposedly. But most of the bugs I've ever heard people complain about are things that were solved a year ago, or more. They just keep getting rehashed. And last time I was lurking through some mailing lists, PostgreSQL had been given a serious speed boost. And the PostgreSQL fans do have a good point in one area: a lot of things Web developers do in code are supposed to be done in the database. But if all you know is MySQL, you're going to become code-heavy when you push MySQL beyond its niche. And some MySQL fans just don't get it, even as they hit the wall.

So there, I've praised and pissed on both databases. What bothers me most about the usual criticisms is how outdated those criticisms are. Try the databases now. They're both kicking serious ass. They're both going to eat into Oracle's markets. Not all of Oracle's markets. But they ARE legit alternatives in some areas.

Re:Ummm....

[Wdomburg]

>Of course, most of the lacking features were
>implemented long ago or are about to see the light
>of day in MySQL 4.1.

I use MySQL in a production environment (not my decision, it was there when I took the job), so I keep a fairly close eye on what developments are being made.

First off, "about to see the light of day" is a bit misleading. The 4.0 branch hasn't been declared stable yet, and still has a few months to go if the 3.23 development cycle is any indication, and it's already approaching two years since the first stable release of 3.23.

There were some major improvements in the 3.23 branch, with the InnoDB table type implementing some of the missing features, specifically row-level locking, transactions, and foreign keys.

There are a few areas where their foreign key implementation flawed though. For instance, you cannot alter a table or create a new index on it without the key constraints being dropped. Likewise, it will allow you to drop a table that another table references with a foreign key constraint. I don't know about you, but I kind of like my database to actually ENFORCE referential integrity.

There are still improvments being made, I'll grant you. They finally added in support for limited ON DELETE triggers (only SET NULL or CASCADE, but its a start).

A fair bit of work is still necessary to add the rest of the features they're missing. Subqueries, at least, are planned in 4.1, as are stable encypted connections and hot backups.

And there are features that are on their "things that have to be done sometime" list, such as stored procedures and triggers, that are crucial to a lot of database configruations. And there are others, like point in time recovery, that aren't even mentioned on their page.

I haven't looked too closely at the query optimizer and analysis tools in 4.0 yet, but the ones in 3.x pale in comparison to PostgreSQL.

(On a side note, I find some of their commentary on how development of MySQL and Postgres differ to be questionable.. They imply heavily that the PostgreSQL team adds features without much planning and without determining an "optimal, definite solution" before doing so, e.g.:

"PostgreSQL is based on a kernel with lots of contributors. In this setup it makes sense to prioritise adding a lot of new features, instead of implementing them optimally, because one can always optimise things later if there arises a need for this".

My impression is almost entirely opposite from following development of both systems, and reading the mailing lists for each. For example, the initial replication code for MySQL was extremely crude and incomplete, whereas the PostgreSQL team has been implementing full WAL and PITR subsystems to base their replication code on.

In fact, there was a discussion pgsql-hackers recently that brought up the question of whether they should START implementing "good enough for now" solutions in certain cases, as opposed to their policy of not including band-aid solutions.

Anyways, enough of my OT rant... :)

Matt

Re:Ummm....

[kcbrown]

I basically agree with the spirit of your statements, but to me it all comes down to this:

If I have to choose between PostgreSQL and MySQL and I don't know ahead of time every detail of the project I'm choosing it for (including the actual future direction of the project -- and how many times have you seen the direction of your project change unexpectedly after you've made all your decisions?), I'll choose PostgreSQL. And the reason is very simple: its capabilities are a superset of MySQL's.

MySQL wins on only one major front, and as you've noted the degree to which it wins on that front seems to be diminishing: speed. There are some minor advantages (built-in full text indexing, if I'm not mistaken) that it has but speed is the reason people claim to prefer it.

I completely agree that you should always choose the right tool for the job. The problem is that the job quite often changes over time. And I can tell you from experience that switching major tools on a production system is not trivial at all.

It's fortunate that the MySQL guys are adding features, but the PostgreSQL guys are doing the same. Both database engines are improving quite rapidly. So there is some chance that future versions of MySQL will have the features you need even if your project requirements change over time. But there's no guarantee of that, so it seems to me that PostgreSQL remains an overall better choice for the typical project when there is any uncertainty about the nature or direction of the project.

Re:Ummm....

[GooberToo]

Um, replication is available both in OS and commercial offerings. I do want to make it clear that only the commercial offering is considered a fairly good solution.

Re:Ummm....

[GooberToo]

MySQL wins on only one major front, and as you've noted the degree to which it wins on that front seems to be diminishing: speed.

That's true as long as you only have a couple or fewer users using your MySQL database concurrently and then, only when they are performing very simple selects. Once you start needing concurrent access, moderate to complex queries, or concurrent select, inserts, updates and/or deletes, MySQL falls far, far behind.

At this point, I'll point back at "my rule of thumb posting", which talks when MySQL should be picked.

Re:Ummm....

[Wee]

But you didn't quite do that, although calling MySQL a DBM file is a bit hostile

Well, I was overgeneralizing, sure. I remember building an used car auction web site in like 1995 where we actually had to think about which "storage system" to use: mSQL, DBM files, or CSV files. When MySQL and the DBI stuff came out later on, I thought I was in heaven. But I've always had it in the back of my head that as long as MySQL is treated like a way to get data via really fast SELECTs, then there should be any problems using it. Not that it can't be used for more than that, don't get me wrong.

So assuming we're being reasonable, here is what each side basically knows (and exaggerates) about the other.

I think you've said it very well, and summed up the two nicely.

And the PostgreSQL fans do have a good point in one area: a lot of things Web developers do in code are supposed to be done in the database.

Well, I've always gone by the theory that if there are lots of people much smarter than me who do things with optimizations in mind, I try not to re-invent that wheel. I'll do stuff in SQL whenever possible. So if I really need something like transactions, then I'm probably better off letting PostgreSQL handle than rather than roll my own in code. It can be done, but there's not very many compelling reasons to do it.

One other thing to consider is the tools and help you get with MySQL over the others. There's lot of code and tools built around MySQL, and that can be a deciding factor as well. And not everyone needs ACID-complinace every time.

So there, I've praised and pissed on both databases. What bothers me most about the usual criticisms is how outdated those criticisms are. Try the databases now. They're both kicking serious ass.

They are indeed. I've used both, and liked both. Having the right tool for the job means having a choice. Each is completely capable in its own right, even for non-web applications. I wouldn't put the DMV or like Farmer's Insurance behind a MySQL database. But would I do a dental office with it? You bet. Choice is good.

-B

Re:Ummm....

[Anthony Boyd]

MySQL wins on only one major front, and as you've noted the degree to which it wins on that front seems to be diminishing: speed.

That's true as long as you only have a couple or fewer users using your MySQL database concurrently and then, only when they are performing very simple selects.

That is not true. It's another old myth that has been disproven.

Re:Ummm....

[Anthony Boyd]

I'm reading these posts and I'm getting the impression that people just don't like Oracle for one reason or another.

I think you're "eavesdropping" on a longstanding developer dispute, and without all the background, it appears to be one thing, when really it is another. My impression is that every database guru respects Oracle. Even if they don't use it, they acknowledge that Oracle works. But the problem is that the PostgreSQL guys take snide jabs at the MySQL guys (right in this thread witness one suggesting that "real" applications use stored procedures, painting any use of MySQL as illegitimate). And the MySQL developers just dumbly hit the ceiling of MySQL's capabilities and keep going -- doing work in code or making kludges. Both sides respect Oracle. Oracle is Switzerland. But neither side respects the other. So when you see comments about both sides taking market from Oracle, that is not about killing Oracle so much as it is trying to find a common ground.

And frankly, I think it is working. I think that mostly sane, clear-headed comments have prevailed here on Slashdot. And that's saying a LOT. I think that both sides need to accept the other as having very legitimate uses. What's more, those uses are different enough that MySQL and PostgreSQL can complement each other. A developer who is willing to use or consider both is a developer who can probably cover just about any conceivable database need. You want a screaming fast Web site that handles huge load? Fine. You want to use Access on the front end but something else on the backend? Fine. You want to build a database to manage a company's finances and process orders? Fine. ALL THREE of those are fully real needs for a database. Oracle can do all three, and it's mostly good at it. PostgreSQL can't give great results for all three scenarios on it's own. MySQL can't either. But if you're willing to consider using both in the areas they're good at, you can do wonderfully.

Re:Ummm....

[GooberToo]

That's not the stable release. So, no, it has not been proven. Fact is, the tables may be turning but until the 4.x series is a known stable release, you're barking up the wrong tree.

Second of all, while it really is a nice graph and all, I didn't really see what and how it's testing. People that don't normally know how to test these types things traditionally generate worthless/meaningless results. While I'm not saying that's the case here, without knowing the details, I immediately doubt them.

Re:Ummm....

[Anthony Boyd]

Second of all, while it really is a nice graph and all, I didn't really see what and how it's testing.

Then you spoke without following the second link.

Just a stab in the dark...

[ebbomega]

But a website with the words "News For Nerds" in its slogan wouldn't be the best place for it.

-Feeding trolls for 10 years running

Best undelete

[jelle]

This may be 'old stuff', becuase the last time I've used novell was years ago in the 3.x and 4.x days... But there is one thing I've never seen since.

Novell has the filesystem with the best undelete I've ever seen. When a file is deleted, it's really just marked 'ready for deletion when necessary' and becomes invisible (sort of hidden), and it's diskspace is marked 'free/unused'.

With a special undelete tool, a user can later undelete any of his files, as long as they haven't been overwritten. And the OS minimizes that. The lower the diskspace utilization, the longer that is. In practice, it's easily more than a couple of days, often weeks.

Re:Best undelete

[evilviper]

Novell has the filesystem with the best undelete I've ever seen.

It's okay if you need to recover a deleted file. If you delete a folder, you've got to do quite a bit more work to get it back.

and it's diskspace is marked 'free/unused'.

Actually, the disk space is still marked as used, but is made available when needed. (as of Netware 5.1)

With a special undelete tool, a user can later undelete any of his file

A 'user' has to have administrative permissions ('S' IIRC) to the folder to undelete a file. To undelete a folder, you have to have Admin priv for the volume.

Besides, Windows has a Recycle Bin, Mac has the Trash, etc. Novell isn't all that great.

There are some cool things about Netware though. If a file has been unused for a while, it will be compressed to save disk space. After it continues to be unused, it will automatically be moved to your archive device (if you have one). So while you see a file on your Netware filesystem, it may actually be on a tape jukebox, and will be restored when you access it.
Did I mention that all this happens AUTOMATICALLY and TRANSPARENTLY?

Re:Best undelete

[larien]

Yup, the HSM module. You can do the same with most operating systems using Veritas/ADSM and various other products. We use a repackaged version of Veritas on Solaris to do the same thing.

Re:Best undelete

[jred]

The disk compression works great, until you either run out of disk space (meaning less than 500mb) or need to move a bunch of older files. Then the entire system slows to a crawl trying to uncompress them.

I especially like how you can tell Netware 4.x *not* to use compression on a volume. It decides you didn't really mean it and compresses files anyway.

Re:Best undelete

[cscx]

FYI, where I worked, we rebooted our Novell boxes waaay more often than I ever rebooted mah 'lil IIS 4 server that happily sat in the corner serving out a couple of websites. Pretty much only after I applied a hotfix did I ever need to reboot.

Re:Best undelete

[cscx]

I'll vouch for the first. The guy who designed and set up the LAN had his head up his ass.

Re:Best undelete

[evilviper]

I like Netware... I just meant that the undelete functionality isn't all that striking.

In fact, one of these replies mentioned the software used for Undelete functionality on Solaris.

Netware is great if you have to admin Windows workstations. Otherwise, I'd much rather be using a Unix system.

Although, saying I'd use Netware rather than Windows as a server isn't saying all that much. It doesn't say much for Netware that it's tools to administrate Windows workstations are far better than Unix's (non-existant) tools.

Re:Best undelete

[Nailer]

There are some cool things about Netware though. If a file has been unused for a while, it will be compressed to save disk space. After it continues to be unused, it will automatically be moved to your archive device (if you have one). So while you see a file on your Netware filesystem, it may actually be on a tape jukebox, and will be restored when you access it.
Did I mention that all this happens AUTOMATICALLY and TRANSPARENTLY?

It certainly is cool, but I do believe Windows 2000 has had this feature since release (I bet the idea was borrowed from Netware though). I'm not sure of the name for it, because I'm mostly a Linux admin these days - does anyone care to fill us in?

Re:Best undelete

[miffo.swe]

Novell has other fetures i like more. The ability to install an application on ONE desktop and then load all register keys onto any computer with windows that uses that program when it loads is a pretty good feture. You install an application once and it then works without hassle on all other machines.

NDS is pretty souped up too and makes AD and LPAD look silly in comparison. It can handle silly amounts of objects in the tree without crumbling. Its enough to drive a whole e-commerse site on.

NDS exists for linux too so interoperability is not an issue. A client for linux would be just what linux needs. My dream network would be Netware on linux and linux clients. A better network to administer cant exist.

The companies that hoose novell and installs AD is in for a bigtime dissapointment because of the extreme lack of fetures in windows like filesystem limitations etc.

Re:Best undelete

[Dwonis]

A 'user' has to have administrative permissions ('S' IIRC) to the folder to undelete a file. To undelete a folder, you have to have Admin priv for the volume

Not in Netware 2.x/3.x. I don't know if this changed later, but I distinctly recall salvaging files from my home directory at school. I did NOT have 'S'upervisory permissions on my home dir.

Re:Best undelete

[jelle]

Recycle bins or trash icons just don't compare for many reasons. One reason is that almost all files that are deleted with a program other than the windows exploer are really deleted and not sent to the recycle bin. You just can't rely on it. Plus least under windows, the recycle bin requires constant user interaction. My laptop often gives me a 'disk space low' balloon next to the clock and then I have to click click click and give the thing attention for 5 minutes so that it can delete the recycle bin and 'temporary internet files', or whatever...

With files in the recycle bin, the amount free diskspace that is reported by the OS does not account for the space that can be freed in the recycle bin. With novell it does.

When you do a 'delete' in the windows command line (cmd.exe), files are deleted, not sent to the recycle bin.

Oh yeah, I forgot about the compression in Novell. That is a nice feature too, but the Linux e2compress patches combined with a little perl script in the crontab can achieve the same on Linux, so it's not unique. I Didn't know about the tape jukebox. Thats seems like a realy nice feature, but the price per gigabyte of tape jukeboxes and the license cost for slots in them just make them a really expensive storage medium.

Re:Best undelete

[drinkypoo]

Besides, Windows has a Recycle Bin, Mac has the Trash, etc. Novell isn't all that great.

The problem with the recycle bin and the trash can is that you put files there and then they sit there and take up space. Now, I don't know jack about how MacOSX handles the trash, but in windows it has some serious problems:

1. The trash is only emptied all at once. When you need more space, you nuke the whole thing with one command, or you have to go into the trash itself and manually delete part of the files.
2. When you are about to run out of disk space, the OS pops up a notification that your disk is about full. The notification tool generally chokes your machine horribly. I take this to be a combination of bad software design and bad filesystem design, but I don't really know. If you don't free up some disk (or let the super slow tool do it) before you run out of disk space, whatever ran out of space will fail. Since most applications don't have crash recovery, or only do crash recovery in large chunks, this will not do you much good. You now must start your render over, or your conversion, or whatever you were doing.
3. Files deleted from the command line are not moved to the trash.

So it sounds like Netware's deletion system is far superior than NT's. And I would guess that MacOSX has the same problem, at least items 1 and 3. 3 would be easy to get around in both windows and MacOSX assuming it has the problem as well; All you have to do to fix it in dos emulation mode is patch the vector table to call some other routine, easy peasy. For other applications you would have to actually change some library to change delete to move-to-recycler-on-same-disk. This still should be cake.

There are some cool things about Netware though. If a file has been unused for a while, it will be compressed to save disk space. After it continues to be unused, it will automatically be moved to your archive device (if you have one). So while you see a file on your Netware filesystem, it may actually be on a tape jukebox, and will be restored when you access it. Did I mention that all this happens AUTOMATICALLY and TRANSPARENTLY?

Even NT (2k and XP) have options to compress unused files. NT4 at least had the option to compress all the files on a filesystem, but I don't remember how automated it was. Moving files to offline storage is nice, though. Windows has some kind of offline storage mechanism but I don't know anything about it, so I won't try to comment further on that.

Re:Best undelete

[evilviper]

Netware is without a doubt a very good server... but it does have it's drawbacks as well.

For one thing, I like the Unix permission model much more. Netware's ACLs are much more difficult to get right, and don't really provide any advantage over Unix permissions with that extra work. In fact, Personal Netware for DOS simplified the permissions to Unix style.

Netware doesn't have the scripting capabilities as Unix does. If I want my Unix machines to all share some file, I simply make a script to copy it from place to place, and do any modifications to the file as needed. Since Netware's configurations are not stored in standard text files, it's much harder to do the same thing.

Well, I suppose the differences between Netware and Unix are like the differences between Photoshop and GIMP. With GIMP/Unix, you have much more fexability, and can pretty much do anything you want to. With Photoshop/Netware, they give you all the settings they think you will want and need, but you are in a tough spot if you want to do things differently than they expect you to.

MySQL can do more than 3M records

[Wee]

I just had a client that needed a database larger than MySQL could capably handle (3 million records.) Since their budget was tight, I went ahead and recommended Postgres. Oh! Whoops! Postgres doesn't run natively on Windows.

I've had MySQL databases with more than 3 million records. Many more, in fact. MySQL works fine if the databases are designed properly. Are you sure that you spec'ed out the job properly? Are you familiar with MySQL? You weren't sure what the client needed before you bid the job? What?

You kinda sound like an MS shill is all...

-B

Re:MySQL can do more than 3M records

[MattRog]

Yes, we all know virtually every RDBMS can store gobs of data. It's the whole 'doing stuff with it' wherein lie the rub.

I'm not saying that MySQL can't 'do stuff' to large amounts of data, but one can go to vBulletin.com and read about customers who routinely have to 'prune' their database of posts/threads (think HardOCP). We're not talking 100 million posts but maybe a mil or two. Now, certainly there are parts of the vBulletin application that just don't scale to 10 million posts. But, it seems MySQL has a bit of a problem dealing with complex JOINs, DISTINCTs, COUNT's (not the vanilla SELECT COUNT( * ) FROM table;) which other RDBMSs do not. It is not terribly hard to see why. In your traditional web apps most queries are of the type "SELECT * FROM myTable WHERE myAutoIncrement = CONSTANT" and so the more 'goofy' queries haven't been optimized as much. This is not idle conjecture but something I have witnessed 'in the field' and have done some rudimentary benchmarks in support of.

So, it is not unthinkable that someone doing a lot of heavy lifting with some moderate amount of data (couple million rows) will hit some brick walls with MySQL. As web applications start to 'grow up' MySQL will address the growing demand as they have started to already (with FK support, row-level locking, replication). The hope is, of course, that MySQL will remain just as easy to use and fast after the new features!

Good stuff!

[shplorb]

Great to see Novell is still alive and kicking, they've really taken a beating over the last decade.

I read the other week that they're cashed up with a billion in the bank or something.

Anyways, I love NetWare - rock solid, efficient and fast. Remember the story here about the NetWare box a uni discovered behind a wall? It had been running for years.

Windows file sharing and its' clones just suck, plain and simple. Don't knock NetWare until you've played around with it and/or seen a network setup properly with it.

NDS rocks hard.

A common (but rather misguided) complaint is that NetWare has crappy multiprocessor support - because one CPU is at 95% utilisation and the other is idle. Ever considered there's no need to use the other CPU(s) if the first isn't maxed out? =]

Now, I don't profess to be an expert on it (I'm not a CNA, CNE or whatever the other one is), but from my experience with using it I just like it, and if you have a network of Windows boxes, use NetWare for file/print serving and whatever else!

Re:transactions

[mbogosian]

Un-clued-in programmers will send a thousand INSERT transactions, instead of a single transaction....

Postgres does have the ability to have those INSERTs be part of a single transaction:

BEGIN;
INSERT (...) INTO table VALUES (...); ...
COMMIT;

All the inserts are now part of one transaction.

I'm unsure if one can make this the default behavior (obviating the need for the BEGIN; statement), like one can with Oracle.

I have a feeling that "un-clued-in" programmers are likely to inadvertently cause all kinds of performance (and other) problems (not limited to the database they're using). Just ask the experienced Oracle user how long it took him/her to learn how queries were specifically interpreted and how one could improve them. Heck, do half of query writers out there even know the concept behind an explain plan or explain query?

Don't Sneer at Netware/Novell

[evilviper]

I don't see why anyone would sneer at Netware. If you've got to administer several Windows machines, Netware is by far the best server for the job.

I don't think I'll hear a single arguement that Windows makes a better server... so what else?

Unix servers for Windows clients don't work very well. For one, MS' native solutions aren't very good, and I haven't seen any client-side programs that can rival the Netware client. It's secure, it integrates nicely, it uses strong encryption (RSA) to encrypt all network traffic, etc.

A Netware server may not be too much like Unix, but it's a hell of a lot better than a Windows Server... and if you've got to have Windows clients, you've got to make some sacrifices.

Netware even has tools to allow Unix compatibility (server-side), so I can't see any reason for an Sys Admin to sneer at Novell.

Re:Don't Sneer at Netware/Novell

[evilviper]

failing upgrades? The only automatic upgrade I've had fail was Norton AV, and I don't think I could fault Netware for that one.

breaking certain network functionality I'd say I've seen this, but in a single, very specific instance. With the client32 installed on Windows NT 4.0, HP's print driver (which includes IP-direct printing) always failed to work.

So, I suppose you'll have to give some more details. I've obviously not had the problems you have.

Re:Don't Sneer at Netware/Novell

[Nailer]

I don't think I'll hear a single arguement that Windows makes a better server... so what else?

Most of the grumbling I used to hear from admins is from unstable Netware clients during the 5.x series, and price - IIRC, Windows was simply cheaper.

These days, the main hassles is Windows 2000 client features that only work with a Windows 2000 server - eg, publishing application to users desktop with group policy.

Re:Don't Sneer at Netware/Novell

[Anonymous Coward]

First let me say that I like Novell. I think NDS is good. Their NOS is reliable and their support is very good.

Now I will detail why they will fail.

Small business will continue to buy Windows servers for its apparent ease of administration and huge application base. They will NOT implement Active Directory unless Microsoft Office requires it. Linux will make huge inroads in to small businesses simply because of cost and administrators are getting more comfortable with it. Novell can't compete with Windows "ease of use" and Linux price. They will be cut off of this market. When RedHat ships an email solution with scheduling and supports it, Novell won't have any advantage left. Why do you think Oracle didn't port 9i to Novell?

Now for mid sized businesses. Companies that have around 500-3000 employees. These companies that are currently on Novell will probably stay for the forseable future, however a bunch of them are evaluating other NOSes. To be honest the ones I see are looking at Win2k and implementing Active Directory. However, because of the economic climate these companies are not doing much spending in the IT world now. Linux is growing in these comapnies simply because of cost, and administrators hate of Microsoft.

Now for large companies 3000+. Can't speak for too many of these, I only know a few, but all of them tend to have some commitee to determine what direction they should go (or their parent company tells them!). Those people tend to lean away from Novell because of Novells "uncertanty" as a company. They tend to play it safe and go with Win2k and Active Directory. However they tend to have every other type of OS, and Linux is making some big inroads for nitch systems. I have NOT seen any of these companies switch off of NT or Unix/Linux to Novell, or give state that Novell is their NOS of choice for the next 5 or so years. Not to say that those companies aren't out there, I just haven't seen them.

What I believe would help save Novell would be for them to slowly abandon NetWare, take RedHat Linux much like Mandrake does, and release a version of it called NetWare X. They would have to port GroupWise to it, and improve their NDS support for it, but they could sell it at the same price NetWare 6 ships today. They could even port their file system to it (Although I bet they could use one of the other journaling file systems). All the development dollars that they currently pay for NetWare could be focused on making their version of Linux great for the mid sized and large businesses. Not to mention that their sofware library would increase by a huge percentage. If they did this then released a "small business package" for around $600.00. They could actually grow that business. It looks like Ray Notra (mis-spelled) may have been smarter than we give him credit for.

Will they do something like this... no. So they will slowly die. At least Steve at Apple gets it.

Steve Michael
smichael@netcapade.net

Re:Don't Sneer at Netware/Novell

[BlowCat]

it uses strong encryption (RSA) to encrypt all network traffic

Using asymmetric key cryptography to encrypt all traffic is a stupid thing to do - it's very hard on CPU. I doubt that Netware does it.

Re:Don't Sneer at Netware/Novell

[shyster]

How about the Novell Client32 for Win2000/XP (IIRC, v4.81...the latest version as of about 3 months ago) screwing up Paradox/BDE finding it's NET directory? That's one I'll never get over. Having to use the MS Client for NW is a real PITA.

Or Client32 for NT/2000 that for some unknown reason never lets the PC join a domain so that you can have dual logins. It says it joins the domain, it says it's a member of the domain...but it's authenticated locally.

Or dual logins with Windows 2000 domain and Netware that take forever to process because Client32 has some insanely long timeout values.

Or the fact that Client32 doesn't really get uninstalled unless you run a special uninstaller program...that's not included in the installation folder.

While we're talking about the installation folder, have you ever tried to replace a NIC in a Win9x machine with Client32 installed? Why doesn't the install keep a copy of the install files on the hard drive so I can just point to it? Why does it refuse to install if I don't tell it where to find the Dutch help files? Why does it have to go into 25 different directories?

Or 550 different options in Client32 properties that seem to be mostly meaningless? Or defaulting to banner pages and broadcast notifications when your print job is done? Or defaulting to installing NDPS and Desktop Management? Or being a 20+ MB download? Or having an option for users not to run the login script? Or taking forever to select the tree and context on login? And god forbid that your server is down, because it'll wait for 5+ minutes with no way to stop it...What the hell is it thinking? Maybe the server will come back up while I wait? Maybe the network is congested so I'll just keep broadcasting? Maybe I should try every networking protocol known to man just in case the server is confused?

Besides those issues....uhhh...Client32 just works, I guess.

Re:Don't Sneer at Netware/Novell

[FatherOfONe]

I don't want to repeat every Item you put on the list, so I will do my best to summerize.

Point 1. You say that Novell is easier to use than Windows NT for most stuff. You point out NWAdmin and Console One.

I agree to a point, but take an average user that knows Windows and have them try and admin a Novell Network. I love nwadmin, and think console one is ok, but it is way different than most people know. Also, you forgot about all the console commands that the user/admin will have to know. NetWare still requires you to load inetcfg and filtcfg for a lot for admin stuff. Almost any monkey could setup a NT server. (Excluding Email)

Point 2. You named one large company that uses Novell and say that their tree (Directory) is the largest in the world.

Great point. Name 10 others! For every one you name there will be 20+ others who are using Windows and or some form of Unix. Also it helps when people on your board of directors also are on Novell's board... My point is that almost nobody is commiting to Novell for the future of their business.

Point 3. It would be bad for Novell to brand a Linux distro, because they couldn't charge for it.

This isn't correct. They could and they know it. Some people within Novell agree with me on this issue and have taken this idea up to "upper management". It has fallen on deaf ears.

Now to add some more points why they should get off NetWare and convert to a Linux Distro.

1. No cost in kernel development.
2. No cost in converting Apache to Novell
3. JVM development would not be needed
4. No money to IBM to port Websphere
5. No more FTP development.
6. They could focus on improving Linux for printing
7. They could offer support and service better than RedHat does.

Even Oracle would now run on "Novell" again. They could market Novell in areas that they NEVER were able to before. I know a lot of shops that would NEVER consider NetWare on their mission critical database servers. They tend to live and die Unix. Well now they could have it, and the LAN administrators could admin it with NDS.

Novell would then have a NOS that most people would consider for something other than file and print. Granted their printing is second to none.

Would these changes make Novell what it was 10 years ago? No! Would it keep them from a slow death? Yes.

Let me ask you one question. If Novell did this, and produced a quality product and service, whould you switch off of them?

Re:Don't Sneer at Netware/Novell

[evilviper]

Client32 for NT/2000 that for some unknown reason never lets the PC join a domain

Why would you want to join a domain when you are running netware? It can dynamically create temporary local users if you configure it to do so.

Client32 has some insanely long timeout values

Okay, it does, but they are prefectly reasonable timeouts for general use. And why are you using Netware and a Windows server?

Why doesn't the install keep a copy of the install files on the hard drive so I can just point to it?

Because it expects you to have the CD. Before 2000, Windows was exactly the same way... Change hardware, you need the CD. Or you can copy the Client32 directory to the hard drive and install from there... Then you always have everything locally.

options in Client32 properties that seem to be mostly meaningless

They are quite useful options actually. I don't particularly like the default settings, but it's easy to have netware automatically push the new settings down upon the first login.

having an option for users not to run the login script?

Again, for some users that is a good feature. For people who you don't want to have that option, just push down that configuration to their machines.

taking forever to select the tree and context on login?

Well, you will notice that SMB has similar problems. If you've got a large and/or busy network, it's going take some time. It's not like changing your tree is a daily routine.

And as soon as you use client32 over a slow link, you will love those defaults. Besides, that's still far less problems than you will see with a Windows Domain.

Re:Don't Sneer at Netware/Novell

[shyster]

Actually, most of the time I have to deal with dual login (domain + Netware) it's for a migration...usually to Windows. =) Other times, a specific program dictates it, or company policy, whatever. Fact is, it's done a good bit and should work a good bit better.

Change hardware = needs files is fine...it obviously needs to load something else from the CD. Novell Client loads THE EXACT SAME FILES it already has. Try it. Just click Details and tell it whatever file it's looking for is in the place it's going to copy to. When Windows does this, it's not that big a deal: C:\WINDOWS and C:\WINDOWS\SYSTEM are about all that you'd need. When Client 32 does it...well, let's just say you better enjoy typing.

Changing the tree isn't a daily routine, but for troubleshooting connection issues, it sure is handy. Of course, staring at a machine that doesn't do anything for 5-10 minutes isn't. FWIW, if Win9x doesn't see a domain controller in about a second after login, it'll immediately let you know that it can't be found.

And I do believe the dynamically created local users is what caused the problem joining the domain in the first place.

What problems do you see in a Windows domain like this? Besides browse issues (pretty much non existant in a Novell network, but mainly because clients aren't listed in Net Neighborhood) I can't think of any significant or pattern of problems....

Re:WTF

[Pathwalker]

Where does one have to go to find normal people?

Outside.

Re:What the heck?!

[g4dget]

Oh! Whoops! Postgres doesn't run natively on Windows. [postgresql.org] This is COMPLETELY unacceptable.

Well, tough, either port it yourself or you buy Microsoft SQL Server, or you switch to Linux. The world doesn't owe you a free database server, much less one for Windows.

If anything, I think too many open source projects are ported to Windows. That eats up a lot of effort, supports Microsoft, and the users would be better off switching to a free OS in the long run anyway.

Re:This surprises me

[einhverfr]

Row sizes are now unlimited. See the Postgres
Limitations info page [postgresql.org] for more info.

Well-- actually a row is limited to 1.6 TB... Of course by the time you need more, they can probably change that again ;)

You are right, though,. PostgreSQL: 7 is really full-featured and powerful. However, I think that the drop column issue is a problem and so I do all my prototyping on MySQL.

OTOH, the extensible types system and Object-Relational system rocks :) Inheritable tables make thinks far more powerful, extensible, and usable. All in all, I really like PostgreSQL.

Re:Actually...

[wd123]

It's just frustrating that the Postgres team decides to port to a platform that is now relatively obscure (Netware) instead of a platform that is one of the top 3 in the database market (Windows.) It seems to be a case of "maybe if we ignore this platform, it will just go away." That attitude is disappointing, especially when it comes from a company that I'd like to support.


Please point out to me where anyone said that the PostgreSQL people actually did the NetWare port? It sounds to me more like Novell did the porting and is packaging it with their system. Also, it's entirely possible that a NetWare port would be a hell of a lot simpler than a Windows port. Windows has no real compatibility with programming in "the rest of the computer world," so why would a bunch of volunteer developers spend time on revmaping the whole application to run on it?

Would you rather they got all hardcore about running on Windows or worked on making the DB itself better on the platforms where it already runs?

Re:What the heck?!

[glenebob]

Your angry at a bunch of hackers because they haven't ported a software package for free? Sheesh! As if you have some god given right to run Postgres on windows or something... Sure it would be nice to have a Windows native port, but come on, if you have such a big problem with it, maybe you should shush up and join the porting effort.

I don't think it was even the core Postgres team that ported to Netware (I'm on the mailing lists, seems like I would have seen something about it). It was probly Novell that did (or is doing) the porting. So go get pissed at Microsoft for not porting Postgres to Windows :-)

Re:What the heck?!

[einhverfr]

Well, Firebird is not considered to be at 1.0 yet, but it should meet your needs. I have done some basic development on it and I like it. But not being at 1.0, I would be a little cautious at this point.

Note also that there are clustering solutions for Interbase/Firebird. Of course all databases have some problems like storing all their field names in upper case by default (Firebird), or making it unnecessarily difficult to drop tables (PostgreSQL) or case sensitive default behaviour ot table names (MySQL) ;)

Re:What the heck?!

[g4dget]

Well, I was careful to say "too many open source projects are ported to Windows", not that open source software should never get ported to Windows.

Large database servers are often run on a dedicated machine, so for something like PostgreSQL, I think there is no need to port it to Windows--bringing up a dedicated database server under Linux is an excellent way to start switching to open source.

Often, the cross-platform compromises of supporting something on both Windows and Linux can be harmful to the software in question, making it more complex or limiting features. I think Apache 2.0's thread support is a good example: it causes a lot of extra work and is arguably completely unnecessary for Linux.

Re:transactions

[madprof]

Easy this one! From the documentation:
"If you don't issue a BEGIN command, then each individual statement has an implicit BEGIN and (if successful) COMMIT wrapped around it."

This is all you need to do and is SQL-compliant of course.

Re:This surprises me

[madprof]

Yes you're wrong. :-) The row size thing was fixed years ago. They've come on in leaps and bounds since version 7.
And in my experience of using it for commercial purposes (3 years and counting) it hasn't actually crashed once.
I'm still waiting to tickle a bug. Not that they aren't there but clearly not that prevalent.

Re:what is netware?

[cscx]

Funny, this article comes complete with a 336x280 ad for Microsoft Small Business Server.

At the school I used to help admin at, we had a Netware 3.11 sever with an uptime around three hundred-something days (!). And you could still type "down" and exit to DOS!

Unfortunately we downed it to upgrade it to NW 5.x. That was the end of that.

MySQL vs. PostgreSQL posts - a lot of them!!!

[haggar]

Gusy, hmmm.. while you're all tied up in a nice flamewar between the two camps... you know that MySQL is ported to NetWare, too, didn't you? It's officially supported, in fact, done by some Novell engineers as far as I know.

Strange that only PostgreSQL got mentioned in the headline.

Re:This surprises me

[ppetru]

PostgreSQL's database size isn't limited by the file size limit of the underlying operating system -- it always splits the database into multiple files. The size limit comes (I think, don't have an URL handy) from a 64 bit integer overflowing.

For those who didn't know

[Kinlan]

There is a new beta period just been announced for the latest version here... It has a lot of feature improvements that the non-postgres fans moan about (i.e column drop)

Re:Is postgresql really enterprise-capable?

[joib]

Availability:

I think there are some 3rd party products implementing various kinds of clustering/HA/failover. For 7.3 (or was it 7.4) they are working to integrate replication into the core.

Scalability:

Well, postgres uses a multi-process model, like say, apache. So in principle it can scale quite well on an SMP system. Regarding clustering, I don't know if the current work on replication includes this or not. I'd guess that when you get replication working correctly, adding clustering is not a big deal. However, the kind of clustering were you have many servers working on the same data, like the oracle9 clustering, is still quite far off, I'd say

Secure data:
postgres can do hot backups, yes. The pg_dump program outputs to standard output, so you can easily integrate it into any normal unix backup scheme with tape robots and whatnot.

Performance:
postgres uses the OS file system, raw devices are not supported. So anything that the OS file system layer supports (e.g. raid) postgres supports. There was some talk about supporting raw devices, but it was decided that it was not worth the effort.

What about "Pervasive"?

[PhilHibbs]

Pervasive Software is an offshoot of Novell, that took btrieve and developed it into a rather good database engine, then stuck an SQL layer on top. I always liked btrieve - it was simple, low level, performed like a rocket, and just sat there and did it's job reliably. Very like Netware, in fact. While Windows NT was drawing pretty pictures on the screen, Netware 3.12 was just sitting in the corner being the best server it could be.

Dropping columns

[chthon]

A few years ago I worked with Oracle and it was only easy to drop a column which was at the end of a row.

If we wanted to drop a column somewhere in the middle, we also had to write a script to query the database minus the deleted column, then delete the old table and rename the new table.

Re:MySQL vs. PostgreSQL posts - a lot of them!!!

[bovinewasteproduct]

Ahhh could be that MySQL is not to be found anywhere on the SDK lists, which list such things as Perl5, PHP, PostgreSQL, Apache, etc?

If it was Novell supported (just having a couple of engineers port something does not make if official/supported;at one time oracle had a FreeBSD port done, but you could not get it outside of oracle) I think it would be listed along with the other supported apps.

I also tried google, but nothing came up on a Novell site about MySQL for NetWare (in the top 20 or 30 that is).

BWP

Doesn't MS SQL Server have row limits? Sybase does

[emil]

I've only used the free Sybase 11.0.3.3 server, but as I remember it has a 1932-byte limit per row. You get a warning if varchars exceed this length, and a hard error if fixed-length columns do.

Sybase and MS SQL Server used to be the same code. Did MS do away with the row limit?

Sigh (rolls eyes)

[buss_error]

Linux aficionados might sneer at an old workhorse like NetWare

And if they did, they would be guilty of what we accuse MS of doing all the time, denigrating a technology without understanding it. Besides, Isn't Unix is older than Novell?

Novell has lots of things done right in it. Self tuning as it runs, stabillity, scale-ability, ease (well, maybe not as easy as Unix) of management, flexabillity.

While improvments could be made to Novell (and Linux), Novell hasn't seemed to completely fallen into the trap of features over stabillity/performance. Although I have to say that GroupWise needs work. Try moving a mail box sometime. Or fixing a broken message database. And it is a major pain that the Admin is dealing with what is basically a black box when it comes to GroupWise.

Re:Sigh (rolls eyes)

[buss_error]

Groupwise 6 fully supports moving users between post offices. Curruptions happens very rarely if aver.

Let's just say my experience doesn't bear out your statement.

"Pervasive"?!?!?!?

[FreeLinux]

Btrieve is the biggest, worst, most awful, satanic, abhorent piece of shit there ever was. And, Pervasive SQL 2000 or whatever it is called this week is still the same old worthless Btrieve piece of shit.

In fact, because they are tied to btrieve applications like Arcserve and Peachtree Accounting and a dozen other specialty apps also SUCK!!!!

Later, when I calm down, I tell you how I really feel.

Re:"Pervasive"?!?!?!?

[mikeboone]

In fact, because they are tied to btrieve applications like Arcserve and Peachtree Accounting and a dozen other specialty apps also SUCK!!!!

This is true! I'm writing an app right now that has to programmatically add invoices into Peachtree's 'database', and it is a complete pain in the ass, all because it's based on Btrieve and not a nice, documented SQL database.

It's only possible because a guy reverse-engineered the database layout (and sells a $379 COM component), but even then there's no error checking...you can easily insert data that will corrupt the whole data file.

Exactly, PostgreSQL is a safer, smarter choice

[alexhmit01]

MySQL is faster when doing simple selects. If you don't have any joins, etc., MySQL is faster. PostgreSQL requires some tuning to get it at a reasonable speed.

HOWEVER, once you get it going, PostgreSQL is pretty slick. If speed is really a problem, get a bigger box. Our main database server for our production system is a dual-1GHz machine with 1 GB RAM. It's not that fast a machine, it was a few grand, and it worked nicely for the past year.

If we outgrow it, we'll go and buy more iron.

It is MUCH cheaper to spend an extra $3000-$5000 on server hardware than it is to have 3-5 programmers spend an extra month on the project to work around MySQL's limitations.

If you ever move beyond trivial database needs, MySQL will kill you. Unless you are really strapped for cash (i.e. this is a hobby site that is going to have enough traffic that performance matters), you're better off going with PostgreSQL. If you find that you need the advanced options, you've got em. If you don't? You'll have 75%-95% of the performance anyways.

Alex

Re:Doesn't MS SQL Server have row limits? Sybase d

[MattRog]

ASE 12.5+ (maybe 12) has the ability to go up to 16K row sizes (with the new larger page sizes) although you would be hard pressed to find a practical instance in which you *need* a row that big -- provided your relations are properly normalized.

The main problem that *existed* in PostreSQL was that TEXT (and other unstructured datatypes) live on-row in PostgreSQL (and MySQL, but I'll get to that later ;)). So you could easily hit the row cap with gobs of text (or a single image!). Now that the cap is gone you shouldn't see any problems. Aside from TEXT you should never hit the limit anyway unless you design horrible tables. :)

Note that TEXT (IMAGE, etc.) data lives off-row in Sybase and MS SQL (and Oracle allows you to specify both I think!). This is very nice since you can keep your rows smaller (better cache hit ratio) and it also allows for a much higher probability that your TEXT data will be in a contiguous block saving expensive disk seeks.

Re: MySQL the Gemini table type has (IIRC) an easily hit max row size if you use text columns. Of course, no one uses Gemini over InnoDB so it is a non-issue.

InnoDB rows live on a HUGE 16KB pages (recompile MySQL to change; yuck!) which means the max row size in older versions is a function of page size (PAGE_SIZE / 2 - STUFF). In later versions this was removed, although limits still exist: http://www.mysql.com/doc/en/InnoDB_restrictions.ht ml But in short, know your data and you should have no problems.

Actually 4.x made it better

[swb]

3.11 was really stable, but it wasn't terribly scalable in terms of the users database. NDS, introduced in 4.x, made it an amazing, seamless enterprise file-print platform. Novell to this day still has the best filesystem ACLs. I learned Novell's first and I was pretty much amazed that anything as bad as UNIX and NT ACLs were even usable by anyone.

My gut feeling is that Novell should have dropped Netware-the-OS and instead ported Netware-file-print-services as a userspace application that could be run on more capable multipurpose operating systems, much the same way of Samba.

What killed Netware wasn't that it didn't do its primary purpose (file-print) better than anything else, but that it was a *horrible* operating system -- 4.xx relied on cooperative multitasking, had no protected memory and couldn't host applications very well and was hard to develop for. The 90s brought a huge surge for quickly written or ported apps that could be bolted onto "the server" -- Netware had few apps and those it had often ran poorly.

Places that would have kept NW file/print but didn't because they needed a more flexible OS could have migrated to a unix flavor and kept running a netware file/print.

Re:Actually 4.x made it better

[Jim Norton]

They've added memory protection since 5.x ... it's configurable too, you can choose which applications are resident in protected memory and which ones you don't. It's great for flakier NLMs that can risk your uptime if they go down.

Re:Actually 4.x made it better

[swb]

With the rub being that some of the worst offenders *cough*Groupwise*cough* were Novell-sourced NLMs...

Yeah! 3.12 baby!

[zerofoo]

I've still got clients running 3.12 as a basic print and file server. Some servers have uptime of well over a year. It might be old an clunky but it still works.

-ted

Re:transactions

[GooberToo]

Your remarks exactly hit the mark. You would be amazed at how often on the mailing lists asks why their inserts are so slow. After being instructed on the proper way of performing multipl inserts within a a single transaction, any remarks on PostgreSQL being slow soon disappear.

Why isn't Novell a SAN or filesystem vendor?

[emil]

It sounds like Novell has enough expertise to make a better vxfs than Veritas, and a better SAN/NAS device than EMC.

I wonder why they never capitalized on these markets.

As a rule of thumb...

[GooberToo]

If you would use an Access database for your project, then MySQL *may* be a good choice for your project. Think of MySQL as a faster, feature poor Access database. Think of Access as a slower MySQL feature rich database. Either case, both stick at concurrent (multi-user) access.

If the answer is no, then you should be looking at using PostgreSQL or a commercial database offering.

Much better number than that...

[Nick Driver]

Linux has, grand total worldwide... abso-freakin-lutely ZERO client licenses!!!!!!

Re:Funny Novell Ad (somewhat OT)

[Software]

I've saw a pretty good Novell ad in a trade rag. The capitalized part looks like a road sign (sorry, I can't do the centering right:

Here's a sign of Microsoft's
new pricing structure.

TOLL
BOOTH AHEAD
HAVE MONEY
READY

Steer clear of costly new licensing "agreements". Head straight for Novell one Net Business Solutions.

I think my favorite part is the the quotes around "agreements".

Re:What the heck?!

[sootman]

Of course all databases have some problems like... making it unnecessarily difficult to drop tables (PostgreSQL)...

Huh?
mydb=# drop table <tablename>;
Works fine for me... I've only been using PGSQL for a year or so, is there something I've been missing? If anything, dropping tables is a bit too easy. ;-)

Re:transactions

[Vlad_the_Inhaler]

Beginners - especially self-taught beginners - will not read that or understand the implications if they accidentally saw it. They might even think that it is something they want. I ran across such a 'programmer' a few years back who was updating enormous tables, one record at a time. This was on a mainframe. We noticed him because the Audits that were being written were filling up twice an hour instead of twice a day, and the discs were being thrashed. He claimed that his update was 'optimally programmed' and that the DB driver was at fault. Changing two lines in his routine reduced the run-time from several days (if we had not killed the run) to around 15 minutes.

Novell will not fall into that trap . .

Re:MySQL vs. PostgreSQL posts - a lot of them!!!

[GooberToo]

For what it's worth, IIRC, the same programmer that has been working with the PostgreSQL guys also mentioned that they've been working on MySQL.

It was also noted that MySQL is being done because it's so well known/prevalent but that PostgreSQL is being worked on because they want a true RDBMS work horse.

Makes perfect sense to me.

Re:Is postgresql really enterprise-capable?

[NerveGas]

Availability: Keep the hardware running, PostgreSQL will keep running. Our PostgreSQL server simply DOESN'T have problems. The only two times it's been shut down in YEARS were for planned hardware upgrades.

steve

Re:Ever heard of SAP DB?

[ckm]

Obviously you can't do research and should not be in the business of recommending software.

If you had done your research, you would have found SAP DB, a high-end GPL database that SAP was built on, that is supported by SAP, runs on WinXX AND has an Oracle emulation layer.

Oh, and it also has nice GUI tools ala MS SQL...

As others have said, if you want Postgres to run on WinXX, either contribute code or $$$.

Chris.

Re:Is postgresql really enterprise-capable?

[NerveGas]

Availability: Keep the hardware running, PostgreSQL will keep running. Our PostgreSQL server simply DOESN'T have problems. The only two times it's been shut down in YEARS were for planned hardware upgrades.

If that isn't enough, there are replication/clustering options available, both free and commercial.

Scalability: PostgreSQL scales much better than the competitors with the number of CPU's.

Secure data: Hot backups work just dandy, and you can send them anywhere you want.

steve

Re:Yeah! 3.12 baby!

[zerofoo]

I guess my clients would be close to that if their power/UPS systems were as reliable.

-ted

Only one question

[cascadingstylesheet]

I don't think I'll hear a single arguement that Windows makes a better server... so what else?

I really want to agree with you ... but ... WHY does it take 10 minutes to log in?

I've worked three places with NT clients connecting to the network via Novell login. Same thing at each place. What is it doing??? For 10 minutes??? At least Windows networking lets me start working sometime before lunch.

Re:Only one question

[evilviper]

The Netware client is nice to the network. If it fails to make a connection, it does not instantly try again, hammering the server and your network.

If it takes a long time, it's quite likely either your server, or your network is overburdened. Could it possibly be that everyone in the company is trying to log-in at the same time? :-)

Re:Only one question

[cascadingstylesheet]

If it takes a long time, it's quite likely either your server, or your network is overburdened. Could it possibly be that everyone in the company is trying to log-in at the same time? :-)

I thought it might be that, but even when I come in at odd hours, or on weekends, it's the same thing.

And at my current employer, when I arrived, most people were on Win95 and connecting with Windows networking (hey what can I say, it's a state agency). We had snappy logins, mostly, and what variation there was did seem to depend on load.

Then we switched to the glorious NT/Novell future (about 4 years late) and I was back in "go get a cup of coffee" mode. Oh well.

Re:Only one question

[evilviper]

All I can say is that I have never even heard of such a problem. Without extensive knowledge of your setup, all I could offer is to go through a list of possiblities.

Maybe your Netware servers are configured incorrectly, maybe the clients are configured incorrectly, maybe routers are configured incorrectly... I really can't say any more than that your experience is not standard, and is obviously the result of a problem with your site.

I have come across many OLD Netware servers (think 100MHz) that have been up and running since their hardware was top of the line, and to this day are serving up authentication, print, and heavy-duty file services.

Re:This surprises me

[sphealey]

I thought Postgres would have too many limitations to be considered a healthy alternative to Oracle.

Novell have historically included a database with Netware for two reasons. (1) To support their internal directory structure. All that NDS information has to go somewhere (2) to enable rapid deployment of midsized apps, typically developed by customers in-house.

I think Novell would tell you that if you have a super-duty database app you should buy Oracle, but they want a product available for "quick hit" applications.

This theory was more applicable in the 1980s-1990s though. Most companies today either buy prepackaged or already know they need big iron.

sPh

Re:PHPBuilder Link Misleading

[Anthony Boyd]

I just completed my own multi-user benchmark tests and the great performance of PostgreSQL vs the not so great performance of MySQL with InnoDB was quite revealing.

Well, since MySQL can keep up with Oracle, are you suggesting that Oracle has "not so great performance" and that PostgreSQL can outperform Oracle?

Re:What the heck?!

[pmz]

...frankly, I am disgusted and disappointed to not have a Free alternative in between those two.

No, you are disgusted and disappointed to not have a free alternative in between those two.

Problem with Oracle: Larry Ellison.

[Futurepower(R)]

I agree with an above commenter that the discussion has been extremely sensible.

There is an issue with Oracle that affects the the choice of a database for a new system: Larry Ellison is widely reputed to be psychologically unbalanced. He is a billionaire who doesn't have to work, and that also affects everything that he does.

If you can use PostgreSQL, then you have the advantage of not dealing with factors that could cause your database system to become much less attractive in the future. Companies like Novell, WordPerfect, Corel, Powerbuilder and many others have been remarkably self-destructive. They were big players 5 years ago, much smaller now. An now, unbelievably, Microsoft seems to be getting ready for a big fall: Windows XP Shows the Direction Microsoft is Going.

Open Source has a BIG advantage that it is not tied to any one person's ego.

Re:Now if they only...

[haggar]

NetWare Web Access? Is that a large rollout/network you're talking about? I have read extatic testimonials and comments about Web Access, but I still have not heard of a larger pilot. Would you be so kind and shed some more light.. if it's not NDA'd?

Re:This surprises me

[einhverfr]

Actually that is the *field* size (which is actually 1GB)! The row size is actually 1.6 TB. However, if you have a 1.6TB row, you should probably redesign your database ;)

Agreed to a point

[einhverfr]

I am a maintainer of a very large open sourced web application framework written in PHP (see my sig). In the beginning of the project we decided to support BOTH MySQL and PostgreSQL. Because we are trying to go for CRM tools, generally feel that PostgreSQL was better suited for that market (hehe we have *emulate* group/roll permissions in MySQL). But our decision to support both was based on the following considerations:

1: There are more web developers that know MySQL than PostgreSQL,so for small businesses, MySQL support is important, and at the time, I was unable to get PostgreSQL to properly initialize under Cygwin on Windows, so we ruled that out.

2: By supporting multiple database backends, we were forced to develop much more extensible application modules, with different levels of database abstractions. Emulating some of PostgreSQL's capabilities in MySQL proved to bew a challenge and so did making PostgreSQL's record set's behave as if they were forward-only. The end result would be easier to port if necessary to Microsoft SQL Server, Oracle or any other reasonably ANSI-complient RDBMS. Although at least one Microsoft engineer has been quoted as saying "Portability is for canoes" (referring to additional QA overhead of portable software), we feel that software designed for portability is in variably better software than single-platform software because the design process is more rigorous.

The end result, I believe, is a quickly maturing project which is both powerful and extensible, and if we only had one open source RDBMS, that would not be the case.

Re:This surprises me

[TheLink]

The drop column was just an annoyance to me, not so much a problem. People lived with that in Oracle for years. If you're prototyping it may not really matter - just rename the column as a dead one, and cleanup when you're finalizing.

And if the table doesn't have too many triggers and other stuff, with Postgresql you can do this (pseudo sql):
begin;
insert into newtable blahblah select from oldtable;
drop oldtable;
rename newtable to oldtable;
test stuff;
commit/rollback;

Yep you can rollback a dropped table and more, try doing that with other databases.

I managed to do that on a production system (I had backups of course, just in case ;) ).

Link.

Re:This surprises me

[einhverfr]

In the prototyping stages, yes. But not once we go into production. But as several people have pointed out, PostgreSQL: 7.3 will have alter table drop column...

Re:Now if they only...

[haggar]

Just to make sure we're talking about the same thing: iFolder, if I recall correctly? That's in NetWare 6.

Re:Best undelete (now NetWare file system)

[evilviper]

best part obout NW ACLs is that fact that a file or directory can't only have one owner and one group permission set.

In theory, ACLs give you more flexability. However, even with the most complex things I've ever done, there's never been a situation where the Unix permissions (simply creating a group to own a set of files, then adding users as needed) haven't been more than adequate, as well as requiring much less inital time setting them up, and requiring a lot less time maintaining them... And by "much less" I really do mean A LOT of time.

SRWCEMFA is useful, a little more flexible than rwx

Hmm, this make take a while:

99.9% of the time, WCEM are all used together, just like a Unix write bit. Partly, that is because of the way many Windows programs work (e.g. Office).

R is the same as the Unix read bit, and F is just like the Unix execute bit (on folders).

So that leaves us with S and A that Unix doesn't really have. Problem is, I don't find them to be all that useful myself. I mean, if you've given someone the right to modify, and erase a file, I don't see why they should be allowed to modify the permissions on that file as well. So, it's something I've never had any use for.

Additionally, (AFAIK) Netware does NOT come with the BSD chflags (or Linux chattr) attributes. Not that they would be commonly used attributes on a Netware server, but they are very good to have in certain situations.

So, I think the Unix RWX permissions are actually just as flexible as Netware's ACLs. And they are quite a bit less time consuming to set and modify.

Re:Now if they only...

[haggar]

Hi Jim,

would you mind if we brought this communication offline? My email is mario@myrealbox..com

mario