1. Anycast doesn't always work well for tcp. 2. Anycast means BGP, which means large blocks of IPs if you don't want to get filtered, which are hard to come by these days. 3. One major benefit of Akamai besides latency is decreased dependency on ISPs often flaky routing decisions; anycast would go the opposite way and increase this.
As the senior systems engineer for a website with points of sale all over the world but datacenters only in the U.S., and a heavy Akamai user, I can tell you that the amount of time for a 301 (requires tcp handshake and http headers) vs the time for DNS is nearly an order of magnitude, so it's a no-brainer to use DNS for this sort of thing.
I know others and I have been saying this up and downthread, but seriously check out configuration management tools like puppet.
(1) is always going to start in Linux with creating your own repo (you can keep it in sync with just rsync, and sync things from your test repo to your production one after they pass testing) and creating RPMs (or.debs, whatever) for any custom software you're using.
Once you've got that in place, you can do (2) and (3) with your configuration management system, which will download new policy when the system comes on-net and enforce it continually even when off-net, just like Group Policy. Because the configuration is all text, you can easily programmatically edit it, keep it in version control, back it up, etc, and configuration management systems are completely object oriented for easy inheritance.
Of course this probably won't stop the maliciously brilliant or totally idiotic, but I've yet to see Group Policy do that either.
I don't understand how this is any different than puppet (or bcfg2/cfengine): change the classes a system is in, and bam! everything changes.
Now obviously puppet is a textual rather than graphical config system, but as someone who has to admin both, I find that much more convenient (can edit with sed, store in svn, etc).
The beauty of Linux storing basically all config in text files, even for desktop environments, is that you don't need some special tool to manage X, you just create a template for the config file and define what set of machines that template should end up on.
We currently use AMANDA in production on about 75 hosts, and I've evaluated, though not deployed, Bacula. I must say that neither really excites me. I'm trying to move from a tape model to a disk model, and I'm left wondering why I want to have "virtual tapes" rather than something a lot more elegant and designed for d2d backup. Oh do I hate labeling and changing tapes, let alone finding them for restores.
Also specifically with AMANDA the permissions model is a total pain (both the host-based network part and the filesystem part) and calling dump on filesystems as a backup method doesn't exactly get you consistent backups of databases, so you have to have some other hotbackup method to get those to disk, which AMANDA then captures.
So while we're continuing to use AMANDA for webservers and other things which don't *need* to be backed up, but where covering your arse is frequently useful, I'm moving to a different model for servers holding critical data: snapshotting the underlying storage with LVM2 or the NetApp, to have guaranteed consistency, then backing up d2d with rdiff-backup which gets me a top-level mirror (easy-as-pie DR restores) with reverse diffs for old data (space efficient) and a very simple way of trashing old information on a schedule if you need the space. For VM servers I don't worry about doing anything inside the VM, and I just snapshot the underlying storage on the host or SAN that the VM sits on. If you're interested in this sort of thing, email me and I'll share my wrapper scripts.
You'll presumably still be stuck with Retrospect for the non-VM Windows (or AMANDA/Bacula--probably Bacula if you're smart), but this is a great method for key databases and VM servers.
You know, everybody keeps saying this, but it doesn't match up with my experience. If you're running a company of ~25 people or less, such that a decent server and quality colocation are likely out, let alone redundant server hardware, then yeah, you're almost certainly going to have better reliability in the cloud.
Over the past four years, however, my systems at companies in the ~100 employee range, with redundant servers in quality colo with offsite backups, but no dedicated DR setup, have averaged more than one full 9 better uptime than Google/Amazon, for pretty similar pricing (to $50/user/year, eg with Zimbra $28/user/yr allowing $22/user/year for admin time, hardware, and colocation fees). Of course if you are only hosting internal corporate apps and thus don't need the services of a full time admin to amortize, then your admin costs will likely be higher than this--but for a full extra 9 of uptime (4-9's vs 3) that could be worth it.
Now you might object that without a dedicated DR site, should my colo fail (which obviously happens, eg when Sun/Craigslist went down and Rackspace also) I'm out at least a day, and that I've just been lucky that hasn't happened to me in the last four years. Probably true, but first of all, there are a ton of major colos in the U.S., and since it's basically front page news when one of them goes down, I'm not at all clear that the average one is down a day every four years. Second, if you look at Amazon's recent 6+ hour S3 outage, etc, cloud computing downtime can quickly add up to the time to bring a colo back on line anyway.
Third, there is an offsetting risk to cloud computing--in the unlikely event that Google really loses my data (Is that more unlikely than a Tier 1 datacenter going down? Unclear.), who's to say that they won't just cut their losses? Certainly consumer gmail accounts have disappeared before without explanation or recourse. Even if my datacenter dies in a fiery explosion, my company has darn near infinite will to make sure that the offsite backups get brought up somewhere sooner or later. Basically, when services are cheap/utility, that means that you tend to get more out of it than it costs, which is awesome, but the flipside is that when the chips are down, it's worth more for you that your data come back than it is for them to fix it.
Now you might think that there has to be something wrong with this anecdote, since Google/Amazon/whomever clearly have more layers of redundancy than I ever will, and so just logically should have better uptime. But there are two problems with this analysis. First, they have way more complexity than I do, and so a problem elsewhere in the system (again, look at the recent S3 bit-corruption outage) at the software layer can quickly propagate across all those layers of hardware redundancy, and obviously their software setup has to be way more complex to cope with all that distributedness in the first place. Second, a lot of issues where the hardware and data are intact can be fixed, worst-case, by someone just rebooting the darn system (like it or not) or restarting the hung daemon, etc. For me, this is just logging into a portal and doing the deed when my pager goes off, which is likely to be a darn high priority and done quickly. For Google or Amazon, as we've seen, the procedure may be basically the same, but takes hours in order for them to get enough hardware back on line (and sufficiently isolated from the stuff that doesn't work) to start offering service again.
So while for small and large businesses the usual Slashdot answers probably suffice easily, I think that at least for tech-savvy firms with other tech needs (the only kind I've worked at, and admittedly a minority), even medium businesses may be far better off insourcing than you seem to think.
Oh, bullets definitely have a sonic boom. That's the "crack" noise you hear. A lot of people think that's the explosion of the powder, but that's false--modern white powder just burns very rapidly, it does not explode. This is why "silenced" weapons not only have "silencers" on the end of the barrel to diffuse the gases so they don't expand into the air so rapidly, they also use special subsonic ammunition.
I can't answer your original question, but for a basic file server, why not just use Linux and lvm2? It handles both snapshots and disk-redundancy without needing hardware RAID. Manifestly production stable, and scales to many terabytes of storage.
You know, I've seen a lot of people complaining about not being able to get XP on new machines, but while Dell's "SOHO" store seems to make that difficult/impossible, the "Small Business" store (which does not require a volume floor, dedicated rep, or anything else) has always made XP fully available. Same goes for the Lenovo website. I fully anticipate both with continue holding out as long as Microsoft lets them in response to market demand from companies like mine which just have no incentive to make the plunge (especially as because XP's been out so long, all of our machines are on it, which is nice, and in addition to the expense anyone who's ever tried to upgrade Windows on 100 laptops knows that's pretty much a non-starter). Also, Dell only offers the better-quality Latitude notebooks through the small business store. Why does anyone even bother with the SOHO store? Is their marketing just that good?
Interestingly enough, while I have concerns about Dell's build quality (had server issues that were 'known upfront' like the Optiplex capacitor issue you mention), in my experience they're waay cheaper than HP (>30% on the 4 and 8 core servers we typically buy) and I've been really impressed by their speed of support and delivery.
No, the delivery speed is nothing like CDW, but you're paying (literally) 50% of what the comparable HP from CDW would be (the 30% figure above is based on quotes from HP direct), and the delivery speed is much faster than what we get from Lenovo Direct on laptops, for instance. On support, I've found that they have competent Linux admins, are aggressive about sending out replacement parts, and even though they only officially support Red Hat and SuSE are in fact completely ok with doing troubleshooting on CentOS and Mandriva systems. Also, when I have had problems the "email my manager at" links that go out on the bottom of every email from a Dell employee are monitored religiously. I've had to complain twice to a manager about something, and both times I got both problem remediation and a nice discount for my time and trouble.
This is on about $50,000 a year volume, so I can't speak to what a smaller business might experience, though obviously we're no Fortune500. Also, and this is I think the real Dell advantage, they actually sell totally configurable systems--you might get the impression from their websites that HP and Lenovo do this, but my experience has been that they will only offer decent discounts and lead times (both pretty expected in a corporate situation) if you're buying a stock or mostly stock SKU.
I actually do see speeds like this. We pay $450/mo for a 3mb guaranteed/100mb burst pipe at our Vericenter cage in Boston, and something like that much for a backup facility in Dallas. We have about 300mb of new data an hour to transfer, which we send in hourly batches, and it usually completes at something like 37mb/s. Now this certainly isn't a guaranteed rate, and I wouldn't want to run an application that depended on that kind of throughput, but it shows you what's possible across Sprint's backbone for a not-crazy amount of money.
My company's premier page lets me configure a D420 with the SSD, Core Duo, 1gb RAM, and CDRW/DVD drive--a typical arrangement for the road warriors who use these things, for $1800. $417 of that is the difference between a 30gb 4200rpm drive and the SSD. And also note that TFA says this is available on D620's as well. I'm pretty excited given the HDD failure rate for our field techs.
Interesting, I tried Fedora Core 6 on a PC at work and gave up after one too many broken packages. As for Mandriva 2007.x, I think it's actually the most stable yet, but only in the One (live cd + installer) editions--I've had bad luck with Free, and nothing to do with multimedia codecs or whatever--it just seems to have little quirks that they don't bother ironing out. Also, if you try to turn on the 3d-desktop stuff with an ATI or Intel card, it barfs pretty unrecoverably, in my experience.
So if anyone from Mandriva is reading this--perhaps give some way to recover from choosing the wrong X.org options on install, and maybe focus on a narrower line of releases so that you can do better QA? But I still think they're way ahead of Kubuntu.
Yeah, urpmi sure isn't apt-get, though I have high hopes for smart as that tool continues to mature. I found that the real power of apt-get though is the Debian repositories themselves, which are massively comprehensive and don't have hardly any broken packages in testing or stable, even if you pin to newer versions of some packages (eg, to get the newest Postgres, which is way better than the 7.x series). My experience with apt-get on Ubuntu was a lot less exciting, as I just kept running into broken packages or over-extensive dependencies that dragged in stuff I didn't want.
Mandrake 8.1 was my first Linux distro, and it's just kept getting better since then, with perhaps the two low spots of 10.0 and 2006.0, both of which very unstable for me--I think the former correllated to bankrupty and the latter to the round of mergers. 2006.0 actually drove me to try Kubuntu (I'm solidly in the KDE camp) which I found very lacking from the perspective of a Mandriva user--difficult to uninstall packages I didn't want (because of the way Kubuntu is really just a package which lists all of the KDE packages as dependencies...), with less good wireless configuration support, a less good partition manager, less good multimedia support, etc.
I am now a full time Linux admin, and while I typically use either RHEL/CentOS or Debian on the server, the few Linux workstations in my company are all running Mandriva. The partitioning tool and hardware support are just the best of any distribution I've tried, and with a quick trip to easyurpmi to set up the external repositories, the userland is the best out there as well. I find PLF way easier to use than all the tricks required to get media codecs and such on Ubuntu.
And I still like it enough that even though I do Linux administration for a living, I still offer free Mandriva email support, which perhaps 10 of you have taken me up on, some of you frequently. Seriously...have a problem, I'll help you out if I can. Nothing against the other distros, but despite its reputation as being for beginners, I haven't found anything about it that's less friendly to experienced admins (for instance, the drak tools don't overwrite hand-edited config files the way SuSE's YaST does). Can anyone tell me what has started the 'less good for experts' tagline, other than that experts don't like to be seen using the distro that all the new users are trying out?
Second the pfsense vote. I am the IT Manager for nTAG Interactive and I ended up moving from our previous combinations of Firebox and Juniper Netscreen systems (depending on location) to pfsense. I'm handling 5 LAN networks and 3 WAN networks on a redundant pair of Dell PE1950's (about as low-end as you can get with a 1950, just one Core 2 Duo, but I chose the 1950 for the hot-swap HDDs and hot-swap power supplies for reliability's sake). Anyway, I'm also running squid transparently which works beautifully. I chose quad-port Intel gigabit PCI-X NICs and I can saturate a LAN-LAN connection (say office to QA network) with only about 20% CPU, which is spectacular when you consider how everyone says you really need hardware routing. Anyway, I am an EXTREMELY satisfied user, thrilled with the beautiful web-gui interface which makes everything drop-dead easy (I'm a Linux admin by background but routing tables are definitely easier to grasp in graphical form). And if I ever decide that I need more networks I can just set pfsense to do VLAN tagging and do the actual routing on my Dell managed switches. Really an amazing system, and if anyone wants to set up something similar do feel free to email me and I'll answer anything I can. Cost me $6k total ($2k per PE1950, $437/ea for the 4 NICs), so not the home-budget sorta thing, but we've now got enterprise-grade routing and firewall in one box which is much easier to manage than the Junipers (I think) and more capable at that.
I'm frankly not sure why Ubuntu took over the popularity crown...I also had a pretty bad experience with Kubuntu and went back to Mandriva, which continue to improve at a rapid rate. I've no idea how it is with GNOME, since I don't use that, but with Mandriva/KDE, this really is 3-4 clicks in a GUI. Try it sometime. And as I note in my sig, I provide free Mandriva email support should something by chance go wrong.
Like most things, I'd suggest you try installing binaries with your distribution, to guarantee that all of the libraries are compatible, etc. But it sounds like you may just have a database permissions issue--does your Trac user have update permissions on the DB?
First there was Kat, which seems to be dead for unknown (personal to the lead developer?) reasons, but is still packaged by eg Mandriva, and is very useful, see its Wikipedia entry. Now its successor is Strigi which acts as KPart and KIO-slave. I don't think anyone's currently packaging it because it's pretty new, but there's no real cost to switching something like a search engine, so use Kat for now if you want it, and switch to Strigi when it becomes available for your distribution.
I love the Plastik theme and the customizability of the KDE toolbars, so to each his own on that front. I think you will find that with KDE-look, you hardly have to spend hours looking for themes if you do want something different, however.
I've tried Crossover for switching my mom from Win98 to Mandriva, and was very pleased with the installation process--everything got up and running smoothly. But when for some reason a Windows application would crash (believe it or not, this happens sometimes in Office 97, which is what she is using), that Windows application would refuse to start thereafter without a fresh install of Crossover and MS Office. Their tech support was no help in the matter (not even a suggestion of where I should look for a locked config file, etc) despite several months of effort, so eventually she ended up with a new box running WinXP. This is a great sadness to me, and I really wish I could've figured out how things could go differently. The situation might be different with a current version of Office, I don't own one so I can't check.
It seems to me painfully obvious that Mandriva 2006's installer is light-years better than Windows, and entirely graphical. I've heard good things about SLED10 as well. Sure these distros may not have the market dominance of RH/Fedora or the coolness of Ubuntu, but they've been around for a long time and polished things to a high sheen. Try them out.
I think this points out another benefit of folding@home over SETI--we'll never know whether those SETI cycles were wasted or not unless one of those computers actually discovers ET. But every hour put into folding@home (I was on the Spymac team when I had a desktop) is actually an incrementally useful result that contributes to those papers that are being published on an ongoing basis. Every protein we come to understand is itself a real breakthrough in biology--many of them appear not only in humans but in other mammals as well. It's the difference between donating to Oxfam and playing the lottery, saying that if you win you'll start your own charity--sensible people generally do the former.
Why do you think that spectrum isn't scarce? Remember that higher frequencies are capable of transmitting more information per channel, but at the cost of shorter range. So there's no need to regulate something like wi-fi, which is high frequency and short range, but even VHF spectrum is pretty crowded with military and public safety users, in addition to FM for radio and TV, and lower parts of the spectrum are extremely valuable due to the ability to transmit long distances and the broad channels needed to get acceptable data throughput. It's true that some of this will be freed up as more services go digital and better yet TCP/IP, but mesh networking is not good for low-latency applications, and there's no indication that this one-time savings will keep us ahead of the increasing demand for bandwidth in the medium-term. So bandwidth is certainly scarce now, and likely will be so for at least the next 50 years, which is plenty long enough to plan public policy around.
CC isn't a monolithic license, but rather a package of licensing options. CC-sa (share-alike) is essentially the same as GPL, CC-by is roughly the same as the old 4-clause BSD license, etc. Certainly CC-nc (non-commercial) would be inappropriate for inclusion in a Debian system, since many commercial outfits run on Debian. So it just depends on which CC license is used; not all CC licenses are created equal, and certainly not all are GPL compatible or meet the Debian freedom guidelines.
1) Just like in any language, whether a programmer actually follows the MVC structure is mostly up to the programmer. I mean, I suppose you could specify a templating language which is less powerful in order to more strictly enforce things, but that takes language development time, programmer memory to remember two sets of syntax/conventions, and might be an unhelpful restriction for RAD/prototyping, which Rails is often used for. But rails generates models, controllers, and views, so you can hardly blame Rails for people who put app logic in their views. Or did you just think it was app logic because it was written in Ruby and you are used to single-purpose templating languages?
2) Well-written Ruby which uses the C database bindings just isn't slow. My web app, Kiko, has been Slashdotted 3 times now, with up to 20,000 log ons an hour on a single processor box with 512mb of RAM -- and not only didn't go down, it never even got slow for our regular users. Erins.de gets millions of hits per day on just a couple of boxes. It has to be set up properly to take that kind of abuse, but it most definitely does. As for search, there's just no reason to implement full-text search in the scripting language. MySQL has full text search, PostgreSQL has tsearch2 (a set of C libraries), and presumably other mature databases have their own solutions. All Ruby/Rails has to do is call the C code and format the results. This is a textbook case of using the right library for the right job--if you're doing image processing in Python, do you write your own image processor or do you call ImageMagick just like you would in Rails? Seems to me that full text search on a database is the same kind of well-defined problem set, and should be treated the same way.
Slashdot Mirror
Sorry for the late response, but:
1. Anycast doesn't always work well for tcp.
2. Anycast means BGP, which means large blocks of IPs if you don't want to get filtered, which are hard to come by these days.
3. One major benefit of Akamai besides latency is decreased dependency on ISPs often flaky routing decisions; anycast would go the opposite way and increase this.
As the senior systems engineer for a website with points of sale all over the world but datacenters only in the U.S., and a heavy Akamai user, I can tell you that the amount of time for a 301 (requires tcp handshake and http headers) vs the time for DNS is nearly an order of magnitude, so it's a no-brainer to use DNS for this sort of thing.
I know others and I have been saying this up and downthread, but seriously check out configuration management tools like puppet.
(1) is always going to start in Linux with creating your own repo (you can keep it in sync with just rsync, and sync things from your test repo to your production one after they pass testing) and creating RPMs (or .debs, whatever) for any custom software you're using.
Once you've got that in place, you can do (2) and (3) with your configuration management system, which will download new policy when the system comes on-net and enforce it continually even when off-net, just like Group Policy. Because the configuration is all text, you can easily programmatically edit it, keep it in version control, back it up, etc, and configuration management systems are completely object oriented for easy inheritance.
Of course this probably won't stop the maliciously brilliant or totally idiotic, but I've yet to see Group Policy do that either.
I don't understand how this is any different than puppet (or bcfg2/cfengine): change the classes a system is in, and bam! everything changes.
Now obviously puppet is a textual rather than graphical config system, but as someone who has to admin both, I find that much more convenient (can edit with sed, store in svn, etc).
The beauty of Linux storing basically all config in text files, even for desktop environments, is that you don't need some special tool to manage X, you just create a template for the config file and define what set of machines that template should end up on.
We currently use AMANDA in production on about 75 hosts, and I've evaluated, though not deployed, Bacula. I must say that neither really excites me. I'm trying to move from a tape model to a disk model, and I'm left wondering why I want to have "virtual tapes" rather than something a lot more elegant and designed for d2d backup. Oh do I hate labeling and changing tapes, let alone finding them for restores.
Also specifically with AMANDA the permissions model is a total pain (both the host-based network part and the filesystem part) and calling dump on filesystems as a backup method doesn't exactly get you consistent backups of databases, so you have to have some other hotbackup method to get those to disk, which AMANDA then captures.
So while we're continuing to use AMANDA for webservers and other things which don't *need* to be backed up, but where covering your arse is frequently useful, I'm moving to a different model for servers holding critical data: snapshotting the underlying storage with LVM2 or the NetApp, to have guaranteed consistency, then backing up d2d with rdiff-backup which gets me a top-level mirror (easy-as-pie DR restores) with reverse diffs for old data (space efficient) and a very simple way of trashing old information on a schedule if you need the space. For VM servers I don't worry about doing anything inside the VM, and I just snapshot the underlying storage on the host or SAN that the VM sits on. If you're interested in this sort of thing, email me and I'll share my wrapper scripts.
You'll presumably still be stuck with Retrospect for the non-VM Windows (or AMANDA/Bacula--probably Bacula if you're smart), but this is a great method for key databases and VM servers.
You know, everybody keeps saying this, but it doesn't match up with my experience. If you're running a company of ~25 people or less, such that a decent server and quality colocation are likely out, let alone redundant server hardware, then yeah, you're almost certainly going to have better reliability in the cloud.
Over the past four years, however, my systems at companies in the ~100 employee range, with redundant servers in quality colo with offsite backups, but no dedicated DR setup, have averaged more than one full 9 better uptime than Google/Amazon, for pretty similar pricing (to $50/user/year, eg with Zimbra $28/user/yr allowing $22/user/year for admin time, hardware, and colocation fees). Of course if you are only hosting internal corporate apps and thus don't need the services of a full time admin to amortize, then your admin costs will likely be higher than this--but for a full extra 9 of uptime (4-9's vs 3) that could be worth it.
Now you might object that without a dedicated DR site, should my colo fail (which obviously happens, eg when Sun/Craigslist went down and Rackspace also) I'm out at least a day, and that I've just been lucky that hasn't happened to me in the last four years. Probably true, but first of all, there are a ton of major colos in the U.S., and since it's basically front page news when one of them goes down, I'm not at all clear that the average one is down a day every four years. Second, if you look at Amazon's recent 6+ hour S3 outage, etc, cloud computing downtime can quickly add up to the time to bring a colo back on line anyway.
Third, there is an offsetting risk to cloud computing--in the unlikely event that Google really loses my data (Is that more unlikely than a Tier 1 datacenter going down? Unclear.), who's to say that they won't just cut their losses? Certainly consumer gmail accounts have disappeared before without explanation or recourse. Even if my datacenter dies in a fiery explosion, my company has darn near infinite will to make sure that the offsite backups get brought up somewhere sooner or later. Basically, when services are cheap/utility, that means that you tend to get more out of it than it costs, which is awesome, but the flipside is that when the chips are down, it's worth more for you that your data come back than it is for them to fix it.
Now you might think that there has to be something wrong with this anecdote, since Google/Amazon/whomever clearly have more layers of redundancy than I ever will, and so just logically should have better uptime. But there are two problems with this analysis. First, they have way more complexity than I do, and so a problem elsewhere in the system (again, look at the recent S3 bit-corruption outage) at the software layer can quickly propagate across all those layers of hardware redundancy, and obviously their software setup has to be way more complex to cope with all that distributedness in the first place. Second, a lot of issues where the hardware and data are intact can be fixed, worst-case, by someone just rebooting the darn system (like it or not) or restarting the hung daemon, etc. For me, this is just logging into a portal and doing the deed when my pager goes off, which is likely to be a darn high priority and done quickly. For Google or Amazon, as we've seen, the procedure may be basically the same, but takes hours in order for them to get enough hardware back on line (and sufficiently isolated from the stuff that doesn't work) to start offering service again.
So while for small and large businesses the usual Slashdot answers probably suffice easily, I think that at least for tech-savvy firms with other tech needs (the only kind I've worked at, and admittedly a minority), even medium businesses may be far better off insourcing than you seem to think.
Oh, bullets definitely have a sonic boom. That's the "crack" noise you hear. A lot of people think that's the explosion of the powder, but that's false--modern white powder just burns very rapidly, it does not explode. This is why "silenced" weapons not only have "silencers" on the end of the barrel to diffuse the gases so they don't expand into the air so rapidly, they also use special subsonic ammunition.
I can't answer your original question, but for a basic file server, why not just use Linux and lvm2? It handles both snapshots and disk-redundancy without needing hardware RAID. Manifestly production stable, and scales to many terabytes of storage.
You know, I've seen a lot of people complaining about not being able to get XP on new machines, but while Dell's "SOHO" store seems to make that difficult/impossible, the "Small Business" store (which does not require a volume floor, dedicated rep, or anything else) has always made XP fully available. Same goes for the Lenovo website. I fully anticipate both with continue holding out as long as Microsoft lets them in response to market demand from companies like mine which just have no incentive to make the plunge (especially as because XP's been out so long, all of our machines are on it, which is nice, and in addition to the expense anyone who's ever tried to upgrade Windows on 100 laptops knows that's pretty much a non-starter). Also, Dell only offers the better-quality Latitude notebooks through the small business store. Why does anyone even bother with the SOHO store? Is their marketing just that good?
Interestingly enough, while I have concerns about Dell's build quality (had server issues that were 'known upfront' like the Optiplex capacitor issue you mention), in my experience they're waay cheaper than HP (>30% on the 4 and 8 core servers we typically buy) and I've been really impressed by their speed of support and delivery.
No, the delivery speed is nothing like CDW, but you're paying (literally) 50% of what the comparable HP from CDW would be (the 30% figure above is based on quotes from HP direct), and the delivery speed is much faster than what we get from Lenovo Direct on laptops, for instance. On support, I've found that they have competent Linux admins, are aggressive about sending out replacement parts, and even though they only officially support Red Hat and SuSE are in fact completely ok with doing troubleshooting on CentOS and Mandriva systems. Also, when I have had problems the "email my manager at" links that go out on the bottom of every email from a Dell employee are monitored religiously. I've had to complain twice to a manager about something, and both times I got both problem remediation and a nice discount for my time and trouble.
This is on about $50,000 a year volume, so I can't speak to what a smaller business might experience, though obviously we're no Fortune500. Also, and this is I think the real Dell advantage, they actually sell totally configurable systems--you might get the impression from their websites that HP and Lenovo do this, but my experience has been that they will only offer decent discounts and lead times (both pretty expected in a corporate situation) if you're buying a stock or mostly stock SKU.
I actually do see speeds like this. We pay $450/mo for a 3mb guaranteed/100mb burst pipe at our Vericenter cage in Boston, and something like that much for a backup facility in Dallas. We have about 300mb of new data an hour to transfer, which we send in hourly batches, and it usually completes at something like 37mb/s. Now this certainly isn't a guaranteed rate, and I wouldn't want to run an application that depended on that kind of throughput, but it shows you what's possible across Sprint's backbone for a not-crazy amount of money.
My company's premier page lets me configure a D420 with the SSD, Core Duo, 1gb RAM, and CDRW/DVD drive--a typical arrangement for the road warriors who use these things, for $1800. $417 of that is the difference between a 30gb 4200rpm drive and the SSD. And also note that TFA says this is available on D620's as well. I'm pretty excited given the HDD failure rate for our field techs.
Interesting, I tried Fedora Core 6 on a PC at work and gave up after one too many broken packages. As for Mandriva 2007.x, I think it's actually the most stable yet, but only in the One (live cd + installer) editions--I've had bad luck with Free, and nothing to do with multimedia codecs or whatever--it just seems to have little quirks that they don't bother ironing out. Also, if you try to turn on the 3d-desktop stuff with an ATI or Intel card, it barfs pretty unrecoverably, in my experience.
So if anyone from Mandriva is reading this--perhaps give some way to recover from choosing the wrong X.org options on install, and maybe focus on a narrower line of releases so that you can do better QA? But I still think they're way ahead of Kubuntu.
Yeah, urpmi sure isn't apt-get, though I have high hopes for smart as that tool continues to mature. I found that the real power of apt-get though is the Debian repositories themselves, which are massively comprehensive and don't have hardly any broken packages in testing or stable, even if you pin to newer versions of some packages (eg, to get the newest Postgres, which is way better than the 7.x series). My experience with apt-get on Ubuntu was a lot less exciting, as I just kept running into broken packages or over-extensive dependencies that dragged in stuff I didn't want.
Mandrake 8.1 was my first Linux distro, and it's just kept getting better since then, with perhaps the two low spots of 10.0 and 2006.0, both of which very unstable for me--I think the former correllated to bankrupty and the latter to the round of mergers. 2006.0 actually drove me to try Kubuntu (I'm solidly in the KDE camp) which I found very lacking from the perspective of a Mandriva user--difficult to uninstall packages I didn't want (because of the way Kubuntu is really just a package which lists all of the KDE packages as dependencies...), with less good wireless configuration support, a less good partition manager, less good multimedia support, etc.
I am now a full time Linux admin, and while I typically use either RHEL/CentOS or Debian on the server, the few Linux workstations in my company are all running Mandriva. The partitioning tool and hardware support are just the best of any distribution I've tried, and with a quick trip to easyurpmi to set up the external repositories, the userland is the best out there as well. I find PLF way easier to use than all the tricks required to get media codecs and such on Ubuntu.
And I still like it enough that even though I do Linux administration for a living, I still offer free Mandriva email support, which perhaps 10 of you have taken me up on, some of you frequently. Seriously...have a problem, I'll help you out if I can. Nothing against the other distros, but despite its reputation as being for beginners, I haven't found anything about it that's less friendly to experienced admins (for instance, the drak tools don't overwrite hand-edited config files the way SuSE's YaST does). Can anyone tell me what has started the 'less good for experts' tagline, other than that experts don't like to be seen using the distro that all the new users are trying out?
Second the pfsense vote. I am the IT Manager for nTAG Interactive and I ended up moving from our previous combinations of Firebox and Juniper Netscreen systems (depending on location) to pfsense. I'm handling 5 LAN networks and 3 WAN networks on a redundant pair of Dell PE1950's (about as low-end as you can get with a 1950, just one Core 2 Duo, but I chose the 1950 for the hot-swap HDDs and hot-swap power supplies for reliability's sake). Anyway, I'm also running squid transparently which works beautifully. I chose quad-port Intel gigabit PCI-X NICs and I can saturate a LAN-LAN connection (say office to QA network) with only about 20% CPU, which is spectacular when you consider how everyone says you really need hardware routing. Anyway, I am an EXTREMELY satisfied user, thrilled with the beautiful web-gui interface which makes everything drop-dead easy (I'm a Linux admin by background but routing tables are definitely easier to grasp in graphical form). And if I ever decide that I need more networks I can just set pfsense to do VLAN tagging and do the actual routing on my Dell managed switches. Really an amazing system, and if anyone wants to set up something similar do feel free to email me and I'll answer anything I can. Cost me $6k total ($2k per PE1950, $437/ea for the 4 NICs), so not the home-budget sorta thing, but we've now got enterprise-grade routing and firewall in one box which is much easier to manage than the Junipers (I think) and more capable at that.
I'm frankly not sure why Ubuntu took over the popularity crown...I also had a pretty bad experience with Kubuntu and went back to Mandriva, which continue to improve at a rapid rate. I've no idea how it is with GNOME, since I don't use that, but with Mandriva/KDE, this really is 3-4 clicks in a GUI. Try it sometime. And as I note in my sig, I provide free Mandriva email support should something by chance go wrong.
Like most things, I'd suggest you try installing binaries with your distribution, to guarantee that all of the libraries are compatible, etc. But it sounds like you may just have a database permissions issue--does your Trac user have update permissions on the DB?
First there was Kat, which seems to be dead for unknown (personal to the lead developer?) reasons, but is still packaged by eg Mandriva, and is very useful, see its Wikipedia entry. Now its successor is Strigi which acts as KPart and KIO-slave. I don't think anyone's currently packaging it because it's pretty new, but there's no real cost to switching something like a search engine, so use Kat for now if you want it, and switch to Strigi when it becomes available for your distribution. I love the Plastik theme and the customizability of the KDE toolbars, so to each his own on that front. I think you will find that with KDE-look, you hardly have to spend hours looking for themes if you do want something different, however.
I've tried Crossover for switching my mom from Win98 to Mandriva, and was very pleased with the installation process--everything got up and running smoothly. But when for some reason a Windows application would crash (believe it or not, this happens sometimes in Office 97, which is what she is using), that Windows application would refuse to start thereafter without a fresh install of Crossover and MS Office. Their tech support was no help in the matter (not even a suggestion of where I should look for a locked config file, etc) despite several months of effort, so eventually she ended up with a new box running WinXP. This is a great sadness to me, and I really wish I could've figured out how things could go differently. The situation might be different with a current version of Office, I don't own one so I can't check.
It seems to me painfully obvious that Mandriva 2006's installer is light-years better than Windows, and entirely graphical. I've heard good things about SLED10 as well. Sure these distros may not have the market dominance of RH/Fedora or the coolness of Ubuntu, but they've been around for a long time and polished things to a high sheen. Try them out.
I think this points out another benefit of folding@home over SETI--we'll never know whether those SETI cycles were wasted or not unless one of those computers actually discovers ET. But every hour put into folding@home (I was on the Spymac team when I had a desktop) is actually an incrementally useful result that contributes to those papers that are being published on an ongoing basis. Every protein we come to understand is itself a real breakthrough in biology--many of them appear not only in humans but in other mammals as well. It's the difference between donating to Oxfam and playing the lottery, saying that if you win you'll start your own charity--sensible people generally do the former.
Why do you think that spectrum isn't scarce? Remember that higher frequencies are capable of transmitting more information per channel, but at the cost of shorter range. So there's no need to regulate something like wi-fi, which is high frequency and short range, but even VHF spectrum is pretty crowded with military and public safety users, in addition to FM for radio and TV, and lower parts of the spectrum are extremely valuable due to the ability to transmit long distances and the broad channels needed to get acceptable data throughput. It's true that some of this will be freed up as more services go digital and better yet TCP/IP, but mesh networking is not good for low-latency applications, and there's no indication that this one-time savings will keep us ahead of the increasing demand for bandwidth in the medium-term. So bandwidth is certainly scarce now, and likely will be so for at least the next 50 years, which is plenty long enough to plan public policy around.
CC isn't a monolithic license, but rather a package of licensing options. CC-sa (share-alike) is essentially the same as GPL, CC-by is roughly the same as the old 4-clause BSD license, etc. Certainly CC-nc (non-commercial) would be inappropriate for inclusion in a Debian system, since many commercial outfits run on Debian. So it just depends on which CC license is used; not all CC licenses are created equal, and certainly not all are GPL compatible or meet the Debian freedom guidelines.
1) Just like in any language, whether a programmer actually follows the MVC structure is mostly up to the programmer. I mean, I suppose you could specify a templating language which is less powerful in order to more strictly enforce things, but that takes language development time, programmer memory to remember two sets of syntax/conventions, and might be an unhelpful restriction for RAD/prototyping, which Rails is often used for. But rails generates models, controllers, and views, so you can hardly blame Rails for people who put app logic in their views. Or did you just think it was app logic because it was written in Ruby and you are used to single-purpose templating languages?
2) Well-written Ruby which uses the C database bindings just isn't slow. My web app, Kiko, has been Slashdotted 3 times now, with up to 20,000 log ons an hour on a single processor box with 512mb of RAM -- and not only didn't go down, it never even got slow for our regular users. Erins.de gets millions of hits per day on just a couple of boxes. It has to be set up properly to take that kind of abuse, but it most definitely does. As for search, there's just no reason to implement full-text search in the scripting language. MySQL has full text search, PostgreSQL has tsearch2 (a set of C libraries), and presumably other mature databases have their own solutions. All Ruby/Rails has to do is call the C code and format the results. This is a textbook case of using the right library for the right job--if you're doing image processing in Python, do you write your own image processor or do you call ImageMagick just like you would in Rails? Seems to me that full text search on a database is the same kind of well-defined problem set, and should be treated the same way.