httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chuck Murcko <>
Subject Re: require-client functionality
Date Tue, 07 May 1996 17:22:06 GMT
Paul Richards liltingly intones:
> Rob (Hartill) and some others here have a wacky idea about release
> engineering.
> This needs to get sorted out pretty sharpish. The last release was an
> embarrasment since there were numerous serious bugs in it that had to be
> promptly fixed. We promised ourselves that wouldn't happen again but it
> seems no lessons have been learnt :-(
> If we're in beta test then lets do testing and stop developing.
The group can choose to leave out things that are not fixes, or include
them, if they both fix and make more usable an existing feature. This is
especially true if they *are* getting tested on multiple platforms (I
have Solaris, IRIX, FreeBSD, BSDI, and old UnixWare(ugh) here.)

> Apache is getting respectable, people are seeing it's number 1 and
> third party companies are writing code for it. We now have people from
> Cygnus and elsewhere on this list and we need to sharpen up our
> attitude. This is now a world renowned project, we need to stop being a
> bunch of hackers having fun with no regard for the product and start
> behaving like a product development team.  That doesn't mean it won't
> still be fun but there need to be periods of restraint around
> releases.
Whoa there, or we'll have to spin off a company and do an IPO and get
on some magazine cover, barefoot. 8^) Let's just be careful where we
stick our attitude after we sharpen it. 8^)

Don't misunderstand the previous three lines and think I don't understand
real release engineering, please. You're right to say we can improve our
methods as a group. We also need not to lose sight of the group dynamic
that got us here.

> We're supposed to be shaping up a release, that means we test, test,
> test on as many platforms under as many different situations as we can
> to shake out those bugs that casual usage doesn't show up. The tree
> shouldn't be touched other than to fix showstopper bugs.
Showstopper bugs from my viewpoint:

1) Keepalives don't initialize on for virtual hosts (Rob had a user patch
   for this, but it hasn't been committed).

2) Imagemap files must contain absolute URLs, with FQDN, etc. for virtual
   hosts accessed my MSIE and Spry Mosaic. Netscape works OK.

3) Background CGI scripts never close connections, due to some unforseen
   effects in either bclose() or bflush().

4) Apache signal handling is still weird on some OS (IRIX, UnixWare).
   SIGTERM doesn't, occasionally. SIGHUP sometimes ends up in server
   terminating at log rollover time here, under BSDI. Partial work-
   around is the log rotation util in /support, but that doesn't help
   when reconfiguring servers.

5) UnixWare has broken socket library, which requires different options
   to get working.

Aside: No one on this list has or plans to get 2.x Unixware to test on,
       unless Cygnus does. I have to cut test releases, find a user to
       test 'em on, and try to figure out what happened from their

Nonshowstopper bugs:

6) Multiple cookies sent *to* clients by CGI scripts do not appear to work.
   I will look further into this.

7) Status needs better interface. Brian suggested one, and it's ready, if
   there's consensus to include it. Multiple form data items don't work
   with existing status module.

8) Referer log exclusion directives don't work, never have. There's a fix for
   this, already tested.

Other stuff:

9) htdigest.c

10) regexp raging debate

That's only what I came up with off the top of my head, from testing and
wacky release engineering this past week. 8^)

Chuck Murcko	N2K Inc.	Wayne PA
And now, on a lighter note:
Nothing is illegal if one hundred businessmen decide to do it.
		-- Andrew Young

View raw message