Re: about RHL8 including glibc-2.2.93

[Andrew Walrond <andrew at walrond dot org>]

I have closely monitored the development of glibc for quite a while, and 
am entirely happy with the way Ulrich and Red Hat handle the development 
within the GNU/Linux community.

I use RH on my servers, and Gentoo on my desktops. Both see the benefits :)

Andrew Walrond

Ulrich Drepper wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> [ If you must, assume this mail was sent out not by a Red Hat employee
> but instead the official maintainer. ]
>
>
> It has been brought to my attention that once again a bunch of clueless
> people think they have the right and good reason to bash Red Hat for
> using glibc 2.2.93 + patches in the Red Hat Linux 8 release.
>
> Red Hat included this version of the library on my advise.  My advise as
> a Red Hat employee and with my blessing as the official maintainer of
> the package.  This separation of positions actually exists: I am not the
> maintainer of glibc for Red Hat.  The work I do is paid by Red Hat but
> not controlled by them (asm our sales people and various managers who
> would love for this not to be true).
>
> It was possible to make the decision to use the 2.2.93 code because at
> the time the distribution was finalized glibc 2.3 for x86 was already
> finished.  In fact, the engineering work for almost all parts of the
> library were finished.
>
> The fact that this happened in time is no coincident.  Red Hat has
> invested over the last 3-4 months considerable resources to the glibc
> development.  We had 3 people working fulltime on glibc and the related
> projects for this timeframe.  As everybody can ensure her/himself by
> looking at the ChangeLog the vast majority of work in this critical
> phase of the development of 2.3 was done by these 3 people.
>
> And what is more: in the process of preparing the RHL8 release all the
> resources available for testing the distribution (our QA people and beta
> testers) also helped to debug the new glibc release.  This is at least
> as critical as the development itself.  The participation in the
> pre-release testing for glibc has always been extremely weak.  We get
> usually replies and reports from half a dozen people.  Almost nobody is
> installing the newly compiled glibc on a system s/he is using all day,
> every day.  But this would be the only way to find the real problems.
> If you would look back at the release of glibc 2.2 you'd see that right
> after the release a lot more bugs were reported than during the entire
> test cycle.  People are simply not willing to install a test release on
> the system they depend on.  And for most people this is correct, the
> glibc documentation even warns about this.
>
> But this should make it obvious how completely invalid the complaints
> and accusations are.  Without the testing already performed by Red Hat
> and the users we would be many months behind the current state in terms
> of stability and without Red Hat's willingness to invest in the
> development we would probably not have seen glibc 2.3 being released for
> another 6 to 9 months.
>
>
> So:
>
> - - every organization has the possibility to speed up the development of
> any project by assigning enough developers to the project.
> Unfortunately this does not generally happen, at least not for glibc.
> In fact, looking at the ChangeLog it is easy to see that most
> distributions (with the notable exception of SuSE, Debian, and obviously
> Red Hat) do *nothing* at all to improve glibc (at least official version
> used by everybody);
>
> - - the use of the 2.2.93 sources + some patches in RHL8 was done as
> advised by the maintainer; this was one of the requirements formulated
> after similar complaints were made in the past.  The fact that I am also
> a Red Hat employee is irrelevant.  I do not think that anybody has any
> evidence for a claim that I make decisions which favor solely or even
> mostly Red Hat;
>
> - - the use of 2.2.93 was completely save.  The engineering work was
> already finished.  Any incompatibility would in any case have to be
> dealt with since glibc still intends to be binary compatible with glibc 2.0;
>
> - - without the efforts of Red Hat glibc wouldn't be where it is today.
> And all these improvements benefit every user of Linux despite the fact
> that the costs are mostly carried by Red Hat;
>
> - - the RHL8 release does not use glibc 2.3 to differentiate RHL from
> other distributions and make interoperability difficult or impossible.
> glibc 2.3 was used since all development in the last year or so went
> into the glibc 2.3 source base; the glibc 2.2 branch only received
> critical bug fixes.  glibc 2.3 is better overall, fewer bugs, faster,
> and more standard compliance (POSIX as well as LSB).  Especially this
> last point (LSB compliance) completely destroys the line of
> argumentation used.
>
>
> If somebody still choses to be paranoid and see something bad and
> sinister in everything Red Hat does, answer the simple question: what
> else should be done?  I can say that development of glibc happens as
> transparent as possible since I managed to enforce that we do not keep a
> separate internal CVS archive.
>
> If somebody choses not to believe a word I say because I'm part of the
> conspiracy this is her/his right.  But it is also my right to ignore
> such a person and I hope that with these explanations I can get a few
> more people to realize that ignoring these ultra-paranoid or plainly
> stupid people is the right thing to do.
>
> - -- 
> - --------------.                        ,-.            444 Castro Street
> Ulrich Drepper \    ,-----------------'   \ Mountain View, CA 94041 USA
> Red Hat         `--' drepper at redhat.com `---------------------------
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.0.7 (GNU/Linux)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
>
> iD8DBQE9nC382ijCOnn/RHQRAjuKAKDNzaPPuAxXsAROpA63QGu1MF3y7gCfb1Zo
> 5YMcGeHnak4ruoV48OsIw70=
> =itSS
> -----END PGP SIGNATURE-----
>
>