Re: potato late, goals for woody (IMHO)
On 2000-05-01 at 13:08 -0500, Steve Greenland wrote:
> On 01-May-00, 04:58 (CDT), Mike Bilow <email@example.com> wrote:
> > Slink is so old now as to be borderline useless.
> "Useless?" So all those people running web servers, mail servers,
> development projects, file servers, etc. on slink boxes are deluded?
> "Useless for Mike Bilow's needs?" Certainly. Only you can determine
> what you need.
I freely admit that this is exactly the meaning of the phrase I used; what
else could I mean?
> > With its basis of glibc 2.0 instead of 2.1, slink will not run the Sun
> > JDK 1.2.2, PostgreSQL 6.5, and all sorts of other things we need. As
> > a result, we actually have production web servers and database servers
> > running on potato, which is not technically stable. We regard this as
> > extremely undesirable.
> So if we today relabeled the potato dist as "stable", containing exactly
> the same packages as it did yesterday, you'd be ok? If so, forget about
> the label. You seem to think that the quality of the packages you use
> from the potato dist are sufficiently reliable for your needs. (Again,
> only you can make that judgement.) Great! Use them!
This is not strictly a technical problem. It is also a political problem.
We are trying to use Debian to do real work. I support SPI and the Debian
philosophy personally, but my business decisions are necessarily subject
to other criteria. Just because I have a T-shirt with a swirl on it does
not imply that there is no Windows 2000 running here.
We have been using Linux in business situations since 1993. At first, we
only used it for simple tasks such as dial-in PPP serving or Ethernet
routing. Now we use it for much more serious stuff, like SQL and EJB
serving. All of the publicity in the last year or two has made this an
easier thing to explain, but a lot of people would be a great deal happier
if the database server software said "Microsoft" or "IBM" on it. The only
reason we get away with this at all is because of my personal credibility:
I can say we have been using this Linux stuff for years, and it works.
At the same time, I have to cope with people whose main technical
education is reading magazines. When no one had heard of Linux, we had to
explain why were not using Windows NT or Solaris or whatever. Now we have
to explain merely why we are not using Red Hat. I suppose this is an
improvement, and probably even Bob Young would find that funny.
We would never use potato if I did not personally believe that it was
adequately stable. Nevertheless, Debian's release policy pulls the rug
out from under me. I do not enjoy discussions that go generally like
this: "So, I understand you're running our enterprise database system on
an 'unstable' version of Linux?" "Uh, yeah, but we think it works OK."
"If it works OK, why do they call it 'unstable?'"
We have invested substantial time and effort to use Debian precisely
because we regard it as technically superior. For example, we have an
in-house mirror and are autobuilding potato ISO images every morning at
about 7:00am EDT. (It is on the DS-3 and publicly accessible -- except
for non-US components -- at DEBIAN.BILOW.COM via FTP, HTTP, and rsync.)
Caldera once explained that their basic business was to stand up and "take
the bullet" for the system administrator when something went wrong. If it
is against company policy to use free software, Caldera provides a nice
way to write a check to buy Linux. I certainly do not expect Debian to
take any bullets for me, but I do not expect the release policy to paint
the target on me, either.