Distribution support: how long is long enough? [LWN.net]

[This article was contributed by LWN reader Joe 'Zonker' Brockmeier]

Red Hat's recently announced errata policy has drawn some fire from the Linux community for being too stingy. The new policy guarantees that releases will be supported for "at least 12 months from the date of initial release." To look at it another way, it paves the way for Red Hat to end support for products only one year after release. Red Hat's 8.0 release, officially released on September 30 of last year, is slated for retirement on December 31, 2003. Fourteen months is a fairly short life cycle for an operating system, particularly since most companies and users won't be switching to a new release immediately.

An end of life policy isn't new to Linux vendors, though such a short life span is. SuSE announced last year that the company would be retiring releases after two years. Caldera and Mandrake also end support for their products after a few years, though they seem to have no posted policy stating a specific shelf life for the products.

Some have noted that Red Hat may be trying to move users to its "Advanced Server" product. While the latest "consumer" release of Red Hat is being retired at the end of this year, Advanced Server won't be put out to pasture until 2005. Naturally, Red Hat charges much more for the Advanced Server product.

When a company like Microsoft decides to end support for a product, it puts its customers in a fairly unpleasant situation: Be stranded with an unsupported platform that will no longer receive bugfixes and support for new hardware, or buck up the money for upgrades and possibly break support for older applications and face hardware upgrades. Red Hat's customers are in a different position, however, since they possess the full source to their operating system; there's nothing that says that someone else can't maintain a release past Red Hat's expiration date.

Companies that specialize in Linux support (e.g. Tummy.com, others) could provide longer-term support for companies (and individuals who happen to have the cash) for a fee. For that matter, there's no reason a savvy admin couldn't continue to patch a system on their own without official errata from Red Hat. If demand is great enough, Red Hat users might even form a community effort to release errata for older releases, though that might be more effort than simply upgrading to new releases or switching distributions. It will be some time before we see just how well, or how badly, Red Hat's policy change goes over with the Linux Community. It's likely that it will draw little attention until the expiration dates start to approach.

While many Linux users may complain about having to upgrade or scrounge for patches on their own, there is some justification for Red Hat and other vendors to stop supporting older releases. The Open Source development model moves very quickly, making it difficult for a vendor to continue support for a wide variety of packages that may put out many releases a year. Not only does the vendor need to provide updates for each package, they must ensure that the updates don't conflict with or break other packages that may depend on them. For a company struggling to be profitable while still giving away its software, it may make a large difference in the bottom line.

(Log in to post comments)

little practical effect

Posted Jan 2, 2003 5:54 UTC (Thu) by djao (subscriber, #4263) [Link]

Disclaimer: I am a redhat user, but I have no other relationship with the company.

I see absolutely nothing wrong with their errata policy. Would you rather have no guarantee at all, as was the case before, or a guarantee of "at least 12 months" of errata support? If the guarantee was "at most 12 months" then I could see something to complain about, but given a choice between no policy at all and a policy guaranteeing at least 12 months, I'll take the 12 months any day.

The 14 months of projected errata support for redhat 8.0 is actually in line with redhat's historical trends for point-oh releases. I went through the redhat support pages and found the support lifespans of previous redhat versions to be as follows

4.0 - 10 months
5.0 - 16 months
6.0 - 17 months
7.0 - 28 months (assuming March 31, 2003 cutoff date)

The real story here is how redhat 7.0 has been supported for such an amazingly long time. Viewed in this perspective, their new errata policy is actually nothing new at all, as it is merely a codification of what they have already been doing in terms of errata support.

In fact, historically, the only redhat versions that have ever enjoyed protracted errata support have been the final point revisions within each major version (4.2, 5.2, 6.2), and, though your guess is as good as mine, my reading of the new policy is that this pattern will continue. For example, in the very same announcement it is stated that redhat 6.2 will be supported until March 31, 2003, giving it a full 34 months of errata support, which is well in line with what I think a final point revision should have.

limited support duration was: little practical effect

Posted Jan 3, 2003 11:27 UTC (Fri) by akumria (subscriber, #7773) [Link]

The real story here is how redhat 7.0 has been supported for such an amazingly long time

Actually the real story here is the minimal support duration of commerical Linux offerings.

Perhaps you feel that 28 months is a long time. However most companies demand, and obtain, support for around 5 years from their software vendors.

Microsoft released Windows XP on the 21 Oct 2001; and according to their Life cycle roadmap they will have general support for 5 years and potentially 2 further years of hotfix/security support.

In contrast RedHat released RedHat 7,2 on the 22 Oct 2001; and they intend to EOL hotfix and security support in December 2003.

26 months versus 60 months. RedHat, now less than half as good as Microsoft

redhat and Microsoft

Posted Jan 4, 2003 4:17 UTC (Sat) by djao (subscriber, #4263) [Link]

You managed to miss my point completely. There is a whole world of difference between the amount of support I expect for redhat 7.0, and the amount I expect for 7.3.

You are certainly right to complain that redhat's support lifespans pale next to Microsoft's, but I think redhat deserves praise for actually improving their point-oh support level (from 10 to 16 to 17 to 28 months), rather than criticism for not matching Microsoft.

28 months is an amazing amount of support for a point-oh release. Comparing redhat to Microsoft here is useless since Microsoft does not release operating systems every six months like redhat does.

For a proper comparison, you need to use the final point releases (4.2, 5.2, 6.2, 7.3) of redhat, which are IMO the only versions that should ever be considered for production use at all. The support lives of these versions are well above 12 months, and nothing in the new policy changes that.

Distribution support: how long is long enough?

Posted Jan 2, 2003 9:22 UTC (Thu) by beejaybee (guest, #1581) [Link]

1) There is a point which may be being missed. The notice posted by Red Hat indicates the end of support for (at least) Sparc & Alpha systems.

2) There is a difference between bugfix support & vulnerability support. Personally I think vendors have a responsibility to issue patches for critical vulnerabilities (those which could result in a system being rooted, converted into a spam relay or warez host etc) for a lot longer than 14 months.

Naturally there has to be a cutoff point somewhere, and I do appreciate that there is a world of difference between the end of support from an individual linux distributor and the end of support for linux. But I would hope that Red Hat could set a positive example by promising support for critical vulnerabilities for at least a couple of years beyond their recently announced policy.

Debian trends...

Posted Jan 2, 2003 16:23 UTC (Thu) by hazelsct (guest, #3659) [Link]

Speaking of distribution support, it's interesting to note that the all-volunteer Debian project has gone from dropping support for the old release one month after the new (as was done in the slink->potato transition) to something which while not yet well-defined, is considerably longer. Potato, released in 2000, is still supported, five months after the woody release (and two and a half years after its own release!), and since many users still have systems running potato, there are no plans to end support for it.

Like you point out, we can do this because all of the software is free; we also seem to have a sufficient base of dedicated volunteers to do a good job of long-term support.

Distribution support: how long is long enough?

Posted Jan 2, 2003 18:17 UTC (Thu) by NathanBardsley (guest, #6243) [Link]

I think it's important to step back a little further on this issue. First of all, RH installs have traditionally been "easy" to upgrade in place from one release to the next. Not to say it's perfect, but it's not consistently painful. Also, proper partitioning and storing data in a well defined place have been part of running any operating system for a very long time.

Second, while it hasn't really penetrated the public awareness fully, the traditional Red Hat distribution isn't being developed as a server environment any more. Yes, you could and can continue to use it as such, but try getting the same version of Oracle to run on RH 6.2, 7.1, 7.2, etc. It's not the same and sometimes simply doesn't work.

We're starting to see the separation of Linux into the classic server/workstation division and we're even seeing the functional needs stratifying. How many =servers= need modern 3d and audio features?

So RH is now focusing on two different areas: RH Server & RH Desktop. Fine by me, the one I would want years of support has years of support, and the other -- the one that is going to be driven by the latest and greatest features -- has a long standing upgrade path.

Seems pretty reasonable to me, and besides, can anyone really begrudge RH from trying to actually make a little money at this point? It's not like they're being slimey or tricky about any of this. Any one product can't be all things to all people, and RH are making well-balanced decisions.

Distribution support: how long is long enough?

Posted Jan 3, 2003 15:09 UTC (Fri) by minichaz (guest, #630) [Link]

I think the real issue here is how often new releases are made. Surely all the enhancements/fixes in Redhat 8.1 could be released as updates to Redhat 8.0?

While I don't mind being forced to upgrade between major numbers (7.x to 8.x) for example, I think it should be possible to apply point releases (x.0 to x.1) as a patch to an up to date system.

While I have downloaded Redhat 8.0 for free and appreciate that greatly, I do pay for the Redhat Network (RHN) service ($60 US a year for each system, I have 3). If my subscription to RHN overlaps the end of support period for the versions of Redhat I am using and I don't want to upgrade to a supported version just yet will Redhat refund the usused part of my subscription?

On another note... If you like Linux and want a system that just works out of the box then I can highly recommend Redhat 8. I used to be a Redhat hater too. :O)

Time to switch?

Posted Jan 3, 2003 17:02 UTC (Fri) by ssavitzky (subscriber, #2855) [Link]

I've been using RedHat for a long time now, but this may very well be the year I switch. For a long time now, RH has been drifting farther and farther away from a "traditional" Unix that you administer by editing text files, to one that's practically impossible to administer without their GUI tools. Meanwhile, the man pages are getting less and less relevant, because RH doesn't update them according to their configuration changes.

I'm sure I'm not the only one who only reboots a system when there's a power failure or an absolutely necessary hardware upgrade -- my home network still has a news/mail/dialup server running RH5.1, and the file and web servers are 6.2. Upgrades would be painful, to say the least: the structure of /var/spool/mail changed in major ways between 5 and 6, and programs I rely on, like diald and newsx keep getting dropped. The thing that finally drove me over the edge was the discovery that frm, which is part of Elm, got dropped going to 8.0. I was using that!

My typical time between reboots is six months to a year -- heck; at work I've had login sessions lasting 9 months. There's very little point in installing a distro that's not going to have patches available over what I've come to consider its useful life.

No product for small servers

Posted Jan 3, 2003 20:37 UTC (Fri) by skarkkai (subscriber, #4128) [Link]

Redhat's new policy makes it very difficult to use Redhat on small servers. Using standard Redhat releases with only one year of support is quite out of the question. No one in their right minds updates a server of any sort once a year, rarely even every two years.

Using the Advanced Server (AS) product on a small server is not justifiable financially. Let's assume a small server hardware costs $4000. The price of AS is $800 per year per server. If the life time of the server is 5 years, which is pretty ordinary, the AS software ends up costing the same amount as the hardware. For most customers, that's not acceptable. Most customers never need the support services that come with AS, they only want the security updates.

It would be a wise decision on Redhat's part to offer a version of AS that comes with no support, and costs say $200 per year per server. That would be the kind of price that one can justify for any server, and would still provide nice income for Redhat.

small servers shouldn't be using point-oh releases

Posted Jan 4, 2003 3:41 UTC (Sat) by djao (subscriber, #4263) [Link]

No one is suggesting that you run redhat 8.0 on a small server. In fact, I would argue that the only versions of redhat you should ever even consider for a server are the final revisions (4.2, 5.2, 6.2, 7.3) of each major version.

These final revisions have, historically, had much longer than "only one year of support", and I see no reason why the recent announcement would change that.

small servers shouldn't be using point-oh releases

Posted Jan 4, 2003 16:59 UTC (Sat) by skarkkai (subscriber, #4128) [Link]

In an email I got from Redhat's Mark J Cox he listed end of life dates for errata support for some versions:

Red Hat Linux 8.0 (Psyche) December 31, 2003
Red Hat Linux 7.3 (Valhalla) December 31, 2003
Red Hat Linux 7.2 (Enigma) December 31, 2003
Red Hat Linux 7.1 (Seawolf) December 31, 2003
Red Hat Linux 7.0 (Guinness) March 31, 2003
Red Hat Linux 6.2 (Zoot) March 31, 2003

Which gives support lifetime of about 1.5 years for Redhat 7.3.

7.3 lifetime

Posted Jan 4, 2003 17:28 UTC (Sat) by djao (subscriber, #4263) [Link]

Well, I may be wrong, and I will eat my words if I am, but I think redhat 7.3 is the one version out of that list that is likely to have its support life extended past the announced date.