httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Roy T. Fielding" <>
Subject Re: not a good day for a beta
Date Thu, 21 Dec 2000 02:43:06 GMT
On Wed, Dec 20, 2000 at 06:13:55PM -0800, wrote:
> Should I assume this was sent much earlier today?  It came with a bunch of
> other messages about locus not being available.  Roy, did you just send
> this, or are you okay with the beta going ahead?

I sent it out earlier, but the point is that I don't think it makes
sense to tag an untested tree as a beta release, and I have serious
reservations that the current software is ready for a production
environment.  It may be, but I'd like to see it run on our own servers
first.  That doesn't mean see if it compiles and see if it installs
and see if it serves a few files -- it means running it under load
and making sure the memory leaks aren't too bad and that the server
doesn't crash after a few hours of real traffic.

I cannot emphasize enough that the only production machine that I know
about that is running something like HEAD is, and it spent
most of the afternoon in a comatose state.  That is probably unrelated
to the server on the high port, but shouldn't we at least check first?

Actually, I should be more forceful about this comment....

   Calling this thing a beta just because somebody decided that this
   week was a good week to announce a beta is no better than having
   a marketing department determine when our software is ready for

My gut says it isn't ready.  I want at least three people to say that
it is running in a production environment before the release.  I think
it makes more sense to do that testing on HEAD first, before it gets
tagged as beta 1.

This is a byproduct of what I have been requesting ever since 1.1
started with this lame "alpha" and "beta" distinction.  We should
be just numbering the releases as 2.0.0, 2.0.1, 2.0.2, ... and
simply adding the label "alpha" or "beta" to the name of the
tarball based on the amount of testing and confidence in that
specific release -- then we could tag the tree whenever the whole
thing was buildable.  *shrug*

Or maybe I just haven't had enough sleep today.  I just have a hard
time believing that anyone can be confident of what is in HEAD when
nobody has seen a machine run with it without crashing.  That sucks.

And what is really frustrating for me is that I no longer have
access to any production machines for testing.


View raw message