Slashdot Mirror

Triviality has nothing to do with it. It took time to make, hours according to you. So someone charging $20 is not outrageous IMO.

Yes, the price is "arbitrary", as are almost ALL prices. Set the price, watch the market react, adjust accordingly, repeat. How else can anyone possibly come up with a price?

If the author sells one copy, they have lost money (as per your experience). If they sell 5, then they've broken even IF they simply ship the s/w and walk away. However, if they then have to collect the funds (chase down bad orders, give up a cut of the order to auto-pay systems), support the software (id10t problems as well as s/w bugs), market (websites don't get built for free), sell (someone has to accept the orders, ship the goods), and on top of all this come up with a new version, then there is potentially a lot of time put into individual orders, whereas there may be next-to-none put into others.

So, what price should said software sell for? What criteria do you think should be used so as not to make the initial price arbitrary?

Don't confuse F/OSS with commercial software. Though there is a fair number of F/OSS packages that are created under the ESR "bazaar" model, most of the polished packages are actually created under the commercial model. There are the packages made with "enterprise acceptible versions" that they sell (OpenOffice.org vs. StarOffice), and there are those packages that are set up to sell support and/or hardware and/or services. These softwares may be "free", but they most definitely raise monies for their authors (or at least for their major corporate backers).

Does the Linux userspace still suck?

RedHat offers a course which looks very interesting: https://www.redhat.com/training/developer/courses/ rhd236.html

"The gender ratio is pretty extreme, but it's not 100%--there *are* expert female kernel hackers." Like who? ;)

I don't know how fair it is to single people out--they probably just want to get on with their work without being made examples of. But Val Henson (chunkfs), Mingming Cao (ext3/ext4), and Suparna Bhattacharya (aio, various fs hacking) are three that come to mind immediately.

And the fact that you may not know their names if you aren't a kernel developer doesn't mean much. I don't think most people realize how large the Linux kernel project is at this point....

I may be a little biased since I've taught the courses... but...

If what you're looking for is the typical corporate 1 week training course, then IMHO you can't do much better. But! Let's be honest here, what can you get out of a one week course? If it's a serious course, like these are, you're going to get so much information that you'll have trouble staying afloat. You're not going to master these skills in a week. Depending on your programming skills and operating systems background you're looking at months to years before you're comfortable kernel hacking. What these courses can give you is a lot of solid coding examples to build your skills. But don't take my word for it, the code samples are available at ftp://axian.com/pub/RHD_SOLUTIONS. Oh, and of course the Red Hat info: http://www.redhat.com/training/developer/courses/

If you're not in the typical corporate time crunch mode, I'd definitely recommend college courses. Get some general background in OS design while you're at it if you don't already have it. Oh, and LUGs, lots of lugs have groups doing kernel hacking, not a bad place to start there either, plus LUGs don't cost anything!

*sigh* You're probably right. When the Novell deal came up, RedHat was quick to put a really nice article on their site assuring that they'd never do such a thing to their customers.
Funny how times change.
(Well, I just read that link again, and it still gives me some hope; it really does look like a well thought out plan, and not the usual PR fluff you'd expect.)

>Would everybody please read ESR's essay on "Goodbye, 'free software,' 'hello open source'"? Excerpt:

"We suggest that everywhere we as a culture have previously talked about "free software", the label should be changed to "open source". Open-source software. The open-source model. The open source culture. The Debian Open Source Guidelines. (In pitching this to the corporate world I'm also going to be invoking the idea of "peer review" a lot.)"

And here what Alan Cox has to say (to ESR and OpenSource):

That would be because we believe in Free Software and doing the right thing (a practice you appear to have given up on). Maybe it is time the term "open source" also did the decent thing and died out with you.

Source: https://www.redhat.com/archives/fedora-devel-list/ 2007-February/msg01021.html

And of course everyone should read this http://www.gnu.org/philosophy/open-source-misses-t he-point.html (already mentioned from someone else)

It's kind of sad that a company who powers its hundreds of thousands of computers in clusters with a trimmed down RedHat puts Linux second on the list of operating systems to support with its software.

I know Google's just playing the numbers (far more Win users than Linux) but you would think that there would be at least enough respect present for them to develop and release for both platforms in tandem. Google has the resources to do that, it's almost like some sort of 'love yet neglect' relationship that churns out of American movies these days.

I hope Google has bigger plans than slowly rolling out its apps in Linux well after it's put them out in Windows.

RHEL5 shipped March 14th, 2007. Why not compare it's errata?

I wouldn't count any updates released on 3/14 against RHEL5 on it's ship date - It's a perfect example of how OSS works and how fast patches are available. RH wanted to ship a stable version and didn't want to through last-minute patches into the install routine. What's the first thing you do when you install a new OS? You run the tool for online updates. So on day one 19 patches were available for all the bugs that had popped up since the version freeze to produce RHEL5.

Since 3/14, there have been 42 updates to RHEL-WS5. 11 of them have been after the 90-day mark, so that leaves you with 31 defects in the first 90 days of RHEL-WS5. That's also not using the "reduced" method to match feature-for-feature what Vista has.

However, I think the point is still always going to be that you can't have totally bug-free sofware. But it's how fast are bugs found and fixed. That's what Microsoft can't touch. How long do bugs go unreported so someone can take advantage of them on MS OS? Even once reported, how long do they linger? The same is simply not true for any critical bugs found in OSS.

But it is nice to see MS finally taking security seriously. They've only been trying to do that for 5 years with their Trustworthy Computing Initiative. Why not compare Windows 2003 Server stats, since it was released after the Trustworthy Computing Initiative? 6 months showed 38 defects. If you compare RHEL5 with just the same installed features to match WS2003 in 3 more months, I wonder how it will fair?

Of course, Microsoft had the NSA help them with Vista, which proves again that the more eyes you have on the source code, the better ;-p

I'll stick with CentOS myself... all the benefits of RHEL without the support fee costs.

RHEL5 shipped March 14th, 2007. Why not compare it's errata?

I wouldn't count any updates released on 3/14 against RHEL5 on it's ship date - It's a perfect example of how OSS works and how fast patches are available. RH wanted to ship a stable version and didn't want to through last-minute patches into the install routine. What's the first thing you do when you install a new OS? You run the tool for online updates. So on day one 19 patches were available for all the bugs that had popped up since the version freeze to produce RHEL5.

Since 3/14, there have been 42 updates to RHEL-WS5. 11 of them have been after the 90-day mark, so that leaves you with 31 defects in the first 90 days of RHEL-WS5. That's also not using the "reduced" method to match feature-for-feature what Vista has.

However, I think the point is still always going to be that you can't have totally bug-free sofware. But it's how fast are bugs found and fixed. That's what Microsoft can't touch. How long do bugs go unreported so someone can take advantage of them on MS OS? Even once reported, how long do they linger? The same is simply not true for any critical bugs found in OSS.

But it is nice to see MS finally taking security seriously. They've only been trying to do that for 5 years with their Trustworthy Computing Initiative. Why not compare Windows 2003 Server stats, since it was released after the Trustworthy Computing Initiative? 6 months showed 38 defects. If you compare RHEL5 with just the same installed features to match WS2003 in 3 more months, I wonder how it will fair?

Of course, Microsoft had the NSA help them with Vista, which proves again that the more eyes you have on the source code, the better ;-p

I'll stick with CentOS myself... all the benefits of RHEL without the support fee costs.

Really? According to their website, it's $80. Is there another way to download it that's legal?

Even funnier is the fact that Red Hat released replacements to the common TT fonts under a GPL license. The full-hinted versions will be released circa September 2007.

Where the fuck are all the other companies in sponsoring stuff like this?

It looks like you want to create a business with a business model that relies on you controlling access to your source code. If you really want to do that, you shouldn't build your software on top of a Free Software infrastructure. A lot of effort has been put into Free Software licenses to ensure that the software remains free, and trying to work around that is just going to cause you headaches. Use closed-source software instead. That world understands your needs and desires much better and, as a side benefit, won't make any annoying moral judgements against you. Sure, it costs money, but so will your product.

If instead you really want to build a build a business on top of Free Software, you should look at using a business model that is compatible with Free Software. That means no secret source code, but there are lots of other options for making money. Contrary to popular belief, there are quite a few companies out there legitimately making money from Free Software. Look at what they do.

A non-exhaustive list would include making your money off of support, hired development of new features, branding and packaging, access to non-infectious licensed libraries, early access to fixes, or sales of your hardware. The last option sounds promising for you, but this is something you should look into.

It looks like you want to create a business with a business model that relies on you controlling access to your source code. If you really want to do that, you shouldn't build your software on top of a Free Software infrastructure. A lot of effort has been put into Free Software licenses to ensure that the software remains free, and trying to work around that is just going to cause you headaches. Use closed-source software instead. That world understands your needs and desires much better and, as a side benefit, won't make any annoying moral judgements against you. Sure, it costs money, but so will your product.

If instead you really want to build a build a business on top of Free Software, you should look at using a business model that is compatible with Free Software. That means no secret source code, but there are lots of other options for making money. Contrary to popular belief, there are quite a few companies out there legitimately making money from Free Software. Look at what they do.

A non-exhaustive list would include making your money off of support, hired development of new features, branding and packaging, access to non-infectious licensed libraries, early access to fixes, or sales of your hardware. The last option sounds promising for you, but this is something you should look into.

Everyone considering linking statically to anything should read http://people.redhat.com/drepper/no_static_linking .html.

The issues with dynamically loaded modules it particularly important. For example, if your program's statically-linked copy of glibc tries to load an NSS module from a different version... BOOM! If you link to another library that an NSS module that you load is also linked to, but with a different version... BOOM! And so on. This doesn't apply only to glibc, it applies to any library that may load DSOs at runtime, for whatever reason. For example, gtk theme engines, gnome-vfs protocol handlers, gdk image format handlers, input modules...

I tried finding regular, old, telnet the other day. On Vista, nonexistant. On cygwin, nonexistant. On FC6, nonexistant.

Telnet definitely exists on FC6
http://download.fedora.redhat.com/pub/fedora/linux /core/6/i386/os/Fedora/RPMS/telnet-0.17-37.i386.rp m

Undocumented you say? I think not.

Something you may not be thinking about -- understanding where you need your permissions on your systems is something that I'd personally recommend REGARDLESS of SELinux.

I wonder if that is related to this bug: https://bugzilla.redhat.com/bugzilla/show_bug.cgi? id=242111

Core and Extras have been merged into a single repository, so those names no longer exist. But what you are looking for DOES exist. It's all there in the "Everything" version of Fedora. That's an install tree that we provide at (for example):

http://download.fedora.redhat.com/pub/fedora/linux /releases/7/Everything/

The release notes page was already hosed before this hit slashdot. Go here.

Check here.

Why is this modded funny? This is damn straight advice. 5 years ago I had a short term contract to write some QT stuff on Linux. I think it was redhat. I hadn't used Linux before but I read the manual that was in a PDF file and away I went. Most of the stuff you need to know to get around Linux is in the manual. It's not that hard. Like what did people do when the first got their hands on a C64 or Amstrad 6128 or an Amiga 500? They read the manual. How did people learn to use Lotus 123? They read the manual. What is it these days that people can't read instructions? Goddam if everytime I needed to learn something on my job I had to go to a course I would forever be in training. What do I do most of the time? I RTFM. Why is RTFM such a common expression? Because it's good advice. Of course adding a bloody link would help.

Hey I just found a useful link here http://www.redhat.com/docs/manuals/linux now go RTFMs....n00bs! :-P

At lots of locations in North America.

https://www.redhat.com/training/

Except Linux isn't free, when you include support costs and the reality of dealing with the leading distros. As this mail from Red Hat support to me today (regarding the fact that I can't get the bcraid driver to compile against the RHEL 4 U4 -55 errata kernel) shows, the leading Linux vendors have worse CS than MS (who will, at least, reply with something more than a brush off to Windows driver complaints, in my experience).

Quoted mail below (edited to beat the lameness filter on slashdot):

Case Title : Need to compile bcraid for errata kerne -55

Last Update Comment as of 20-MAY-2007 11:21:44 :

Greetings & Welcome to Red Hat Global Support,

I am sorry to let you know that you have only "Production Support Scope of Coverage" and this issue involves an in-compatible third party package. We will not be able to support this issue. I request you to please contact your third party vendor source of this SDK or search this issue in any of the on-line forums or mailing-lists.

Refer - http://www.redhat.com/support/policy/soc/productio n/

I am changing the status of this service request as "Pending Closure", this issue will "auto close" in next 7days. Let us know if there is something else we can help with in this issue.

Lots of extraneous bs redacted, but the gist is: we don't care, thanks for the $595/yr, now go solve it yourself.

[Kmail] was waking up twice a second for no apparent reason...

There are still some applications that wake up for no reason apparently.

Your complaints remind me of 'Why Userspace Sucks', by Dave Jones.

They really used to be kings of FUD but this is laughable, at best. If it wasn't Microsoft, one could almost feel sorry.

They actually prove the point that MS never, ever invented anything and that huge ecconomical success was just due to a ride on top of monopoly wave. That doesn't cut it anymore - Apple and Free Software community are making supperior products already. Windows is in such a mess that MS is not able to make any real improvements without breaking everything and basically starting from scratch. This is the reason to sing 'software as a service' song - Windows is dead and their desktop market share will start shrinking soon. It is 90%+, so yes - wiles, l take some time but they are in trouble.

Oh, Hilf - yes, we still exist. We are here!

Come on guys, is it really news every time Microsoft patches a new security flaw in Windows? When are we going to see the weekly Slashdot articles about the Linux security patches?

https://rhn.redhat.com/errata/RHSA-2007-0338.html
http://support.novell.com/linux/psdb/bydate.html
http://www.debian.org/security/
http://www.slackware.com/security/list.php?l=slack ware-security&y=2007

I'm not big on voting machines, but if we're going to have them, they should be open.

This guy (Alan Dechert) is active in CA and needs your help. I've ponied up some dough; please join me.

He's speaking at the Red Hat Summit today!

Yes, there are *some* people who will not produce information-products (including software, music, images, or what-have-you) if they cannot globally enforce copy restrictions. Agreed. Such people should, IMO, go in to a different line of work. That is perfectly acceptable for a very simple reason: there will be more than enough people who find good reasons to produce such works in the absence of copyright restrictions. Some people will find an alternative means of making money off freely-duplicated works, and others just because they are altruistic.

Remember that people pay good money for their hardware, and copyright restrictions mean they cannot make full use of it. Copyright isn't actually a "freedom" for the person producing a work....people will still be completely free to produce works without copyright restrictions. Copright may be a "benefit" to the producers of a work (though in practice it is not; it is only a benefit to the distributors of the work, but I won't get into that here), but to call it a "freedom" is incorrect. Copyright law is a restriction on freedom to everyone in the world, and if such a freedom is going to be globally sacrificed, there had better be a damn good reason for it.

The only reason you have given is the false premise that without these restrictions, no-one will produce knowledge-products. Not only is this false in theory (since some people will produce stuff for free, and since some people will find ways of making money off knowledge-products in the absence of copyright restrictions), but there are lots of examples of businesses that make money off a free end-product, and of profoundly useful products made without any profit motive. And there are more where those came from.

That last set of links is pretty important. Google gives all of its services away for free, and yet has a market cap of over 100 billion. Not only are there business models built around free products, but they are very profitable and fiercely competitive.

Also check out this and this. Copyright is still there, but it is unenforced upon the consumer. It will be interesting to see how this selective approach to enforcement will pan out.

It is true that a farmer who gives away his crops for free would go broke, and if farmers could not legally force people to pay for their products then there would be no farmers. However, this observation not apply to information products. Information is fundamentally different from physical products, and business models surrounding it wind up taking a different form than traditional business models (a form which includes a free and/or freely redistributable product).

What we are dealing with is a new kind of abundance. Oxygen is an abundant resource, (anyone can get it for free because it just never runs out). Traditional capitalistic wisdom says that it is not possible to build a business around such resources, and further that no one will produce them because of that. Information is also abundant, once it exists (since it can be duplicated at zero cost by anyone). But it is also strangely non-abundant, since it's initial production requires an expenditure of resources. Traditional capitalistic models have a very hard time categorizing it...is it abundant or isn't it? Copyright law is an attempt at forcing it in to the "limited" category so that the traditional models wil

And while I like FreeBSD, I use FreeBSD, and FreeBSD is a friend of mine, the support for Xen as DomU just isn't there.

Instead I'm stuck with RedHat - who can't seem to release a working Xen version.
https://bugzilla.redhat.com/bugzilla/show_bug.cgi? id=234008

'£300k doesn't even buy you any media! A visit to their head office in North Carolina sees the presentation done from a projector on a desk, with bits of cardboard to stop it wobbling'

OK, I can see a theme emerging here in this thread. I like Linux except a) no support ) no software c) company used cardboard. This is meant to be a joke isn't it?

'Subscriptions take the pain out of purchasing software because they provide everything needed in one all-inclusive price'

was: And so it starts (Score:3, it sure does)

Excellent post. Just to add to that. You can test out the OS yourself with VMWare Player, EasyVMX and getting the ISO of the Operating system.

The public key support for eCryptfs can handle more than just public keys. It includes a communication mechanism with a user daemon that can be queried from the kernel on file open events. There is a pluggable key module interface accessible through that daemon. OpenSSL is currently implemented, but there is nothing stopping anyone from writing a module to use GnuPG or any other key management/encryption backend, all in userspace. The module just needs to accept a key signature, and it can perform encryption and decryption based on whatever that signature refers to.

In other news, eCryptfs has recently been given the go-ahead for inclusion into Fedora:

https://bugzilla.redhat.com/bugzilla/show_bug.cgi? id=218556

In the meantime, you can grab all the userspace stuff from the eCryptfs SourceForge site:

http://ecryptfs.sourceforge.net/

OpenLDAP is one such package.
Fedora Directory Server is another.
Heck, you can even use Active Directory *and* Linux.

Microsoft will offer a minimum of 10 years of support for Business and Developer products. Mainstream support for Business and Developer products will be provided for 5 years or for 2 years after the successor product (N+1) is released, whichever is longer. Microsoft will also provide Extended support for the 5 years following Mainstream support or for 2 years after the second successor product (N+2) is released, whichever is longer. Finally, most Business and Developer products will receive at least 10 years of online self-help support.

Consumers get a little less time:

Microsoft will offer Mainstream support for either a minimum of 5 years from the date of a product's general availability, or for 2 years after the successor product (N+1) is released, whichever is longer. Extended support is not offered for Consumer, Hardware, Multimedia, and Microsoft Dynamics products. Products that release new versions annually, such as Microsoft Money, Microsoft Encarta, Microsoft Picture It!, and Microsoft Streets & Trips, will receive a minimum of 3 years of Mainstream support from the product's date of availability. Most products will also receive at least 8 years of online self-help support. Microsoft Xbox games are currently not included in the Support Lifecycle policy.

Ok. Minimum of 5 years. Seems kinda short, I guess. What's Ubuntu's policy?

From their announcement https://lists.ubuntu.com/archives/ubuntu-announce/ 2005-October/000038.html:

Ubuntu is a Linux distribution for your desktop or server, with a
fast and easy install, regular releases, a tight selection of
excellent packages installed by default, every other package you can
imagine available from the network, a commitment to security updates
for 18 months after each release and professional technical support
from many companies around the world.

18 months. Now for the price, that's exceptional, but your argument had nothing to do with price, and everything to do with version upgrades. If updates are your metric for determining whether users are "forced" to upgrade, look no further than the announced support cycle for Ubuntu 5.10.

They looked like they'd gotten better, no doubt. With 6.06, you get 5 years of upgrades--the same minimum guaranteed by Microsoft http://www.ubuntu.com/news/606released:

Ubuntu is freely available, including security updates for five years on servers, with no restrictions on usage and no requirement to purchase support contracts or subscriptions per deployment.

But wait. The 7.04 release of Ubuntu reverts back to 18 months--they say that the 6.06 series was a "long term support release" https://lists.ubuntu.com/archives/ubuntu-announce/ 2007-April/000102.html.

Ubuntu 7.04 will be supported for 18 months on both desktops and servers. Note that 6.06 LTS is a long-term support release, and so users requiring a longer support lifetime may choose to continue using that version rather than upgrade to or install 7.04.

So we're back to 18 months. Microsoft's stated support minimum is more than 3 times longer than Ubuntu's, except for the aberration of Ubuntu 6.06.

So who's 'forced' to upgrade in order to keep support?

I mainly focused on Ubuntu because that's what the person you replied to was talking about. Redhat, arguably the best known Linux vendor, gives their cycle here: http://www.redhat.com/security/updates/errata/ They give you 7 years of

LFS is overrated for learning, cause you basically spend two days compiling stuff.

A lot of bookstores however have very good Linux books, use those as a start and continue with manpages and google.
The Linux Documentation Project is at tldp.org.

It also helps to install different distro's in vmware to play with.
System Administration of Red Hat servers is different from Ubuntu/Debian...

http://www.redhat.com/docs/ is not too bad, but does contain errors :(
(docs.sun.com can also be useful, even for linux)

I don't know why you think NFS doesn't support failover; check out Red Hat Cluster (PDF) or Sun Cluster. You will need a RAID array that has two host ports, such as VTrak E310s, IBM DS3200, HP StorageWorks 500, or Xserve RAID.

I would not suggest cluster file systems such as Lustre for a small installation; they're generally designed to scale up to hundreds or thousands of servers, but not to scale down to a handful.

I don't think that's right at all. I've come across the same question in 2 different companies now and the answer has always been that you must have a RHEL subscription for every machine you have RHEL installed on. In fact have a read of the licencing agreement:
https://www.redhat.com/licenses/rhel_us.html?count ry=buying+a+Red+Hat+Subscription+from+Red+Hat

Read sections 3.1 and 5.1 in particular. In 5.1 they are saying that you must notify them if the number of installed systems exceeds the number of subscriptions you have, and they will bill you for the extra systems etc.:
Client will promptly notify Red Hat if the number of Installed Systems exceeds the number of Installed Systems for which Client has paid the applicable fee. In its notice, Client will include both the number of additional Installed Systems and the date(s) on which such Installed Systems were put into use. Red Hat will invoice Client for the applicable Services for such Installed Systems on a pro-rata basis and Client will pay for such Services in accordance with this Agreement.

This is why Centos is so useful, you can have as many dev/test/uat/whatever machines as you like without having to worry about subscriptions.

https://www.redhat.com/licenses/rhel_us.html?count ry=buying+a+Red+Hat+Subscription+from+Red+Hat

You need to reread section 5 if you have an RHN subscription. You MAY NOT install redhat software on a machine that does not have an RHN subscription and they MAY ask to audit you.

You're just an armchair fanboy who knows nothing about the real state of OSS. You clearly have no idea of who RedHat employs, what they do and the projects RedHat host and support. Ever even seen http://sources.redhat.com? No of course not. Why would you? You're just a user.

A link to a spam blog, with a link to digg? Come on, can you get any lower than this? Why not just link Directly to the actual information?. See how easy that was? Oh wait, I didn't have a million banner ads.

If you're going to download it, get the latest version instead.

The lazy 'article' is just a link to an already out-of-date Endgadget post. Thank ThePopeLayton and Zonk for schilling instead of informing, and then climb the dirtree to find the latest.

http://olpc.download.redhat.com/olpc/streams/sdk/

see also
http://olpc.download.redhat.com/olpc/

If you're going to download it, get the latest version instead.

The lazy 'article' is just a link to an already out-of-date Endgadget post. Thank ThePopeLayton and Zonk for schilling instead of informing, and then climb the dirtree to find the latest.

http://olpc.download.redhat.com/olpc/streams/sdk/

see also
http://olpc.download.redhat.com/olpc/

SRPMs are not updated with security patches or otherwise, however, and that's a good part of what Redhat charges for.

Presumably someone from the CentOS project pays for a Red Hat support plan (or two, or three), and in return is able to download the product (and get the sources as required by the GPL).

Presumably someone from the CentOS project pays for a Red Hat support plan (or two, or three), and in return is able to download the product (and get the sources as required by the GPL).

http://olpc.download.redhat.com/olpc/streams/devel opment/

For the last ten years, I've been developing Taskjitsu, an open source professional services automation system that tracks time sheets and tasks. It is freely available, GPL-licensed, and commercially supported by PKR Internet.

Taskjitsu is at its core a Java web application, layered on top of Tomcat and PostgreSQL. It runs on Windows, Linux, and any other system that can run Java 1.4. We have RPMs available that work with Red Hat Application Server 1.0 and other JPackage 1.6-derived systems.

Ever since Ubuntu Edgy much of the low hanging fruit in speeding up the Ubuntu boot has already been taken. Looking at the bootcharts for my system since then shows remarkably little time when the CPU is idle once the base kernel has finished loading. This means that running anything more in parallel simply won't net me anything (in fact scheduler overhead and disk thrashing may in theory make things slower).

For example, there is an improvement in the time it takes for the clock to appear from "Ubuntu Dapper Flight 3 Default kernel" to "Ubuntu Feisty Herd 5 generic kernel". The Ubuntu folks worked hard to try an eliminate sleeps from their initscripts and when a sleep was unavoidable they would run other parts of the startup process in parallel. They also made changes to Xorg to prevent it (re)reading so much stuff on launch. There was also the introduction of the readahead script which tries to arrange for as much of the boot time reading to be done in one big chunk. Throughput is higher when the disk is only reading and can utilise it's readahead. An attempt is also made to try and request files in the order in which they are laid out on disk (to minimise disk seeks which hurt performance). In Feisty a move was made to using dash instead of bash for scripts because it was smaller and executes scripts faster.

The only things that seem to win me any gain over the default Ubuntu Feisty install are turning off initscripts for services I absolutely won't use (e.g. ipv4 autoconfig via avahi) and reducing the number of restricted binary driver modules being probed (I have long noticed that the only benefit that recompiling the kernel gives to boot speed is that you can simply leave out features not on your computer making the initial kernel startup where it probes for things you might not have (like which software RAID is faster) a shade faster). It is also worth noting that Ubuntu starts X quite early and continues loading services afterwards which means the gain from disabling one of these "after X" services (like CUPS) isn't so noticeable (but might mean your desktop actually starts responding to clicks a bit sooner).

Profiling the boot to try and improve the readahead takes a long time to run - the profile run seems to take three times as long as a regular boot. It could be argued that you will never gain back the extra time you waited on the profile run...

I suspect reducing the boot further will start to need more complicated procedures, perhaps reordering modprobe.conf and reducing the amount of needless reading of files. Eventually you end up having to do the same tricks as Windows/OSX - e.g. working out where the fastest part of the disk is and copying every file needed to boot there, bringing up the network cardafter the desktop has started, periodically defraging bits of the disk, prelinking...