Mailing List Archive

My life in films...
> I'm using a slightly different scheme to number patches than rst, but
> it's not a big deal. The numerical id of the patch in unique accross
> all types (B, E, P). So the first set of patches ar B1 -> Bn, then P(n+1)
> ->P(m), E(m+1) blah blah blah. I've also created a type "O" for other.
> The list is identical to his. I'd like to get submiter email addresses
> attached to these requests also, please speak up if they are yours.

Ok, well I can't see all the patches that rst mentioned in your own list so
can you put me down as being the champion/owner/whatever of (rst's):

B5) XBITHACK not honored on <!--#include-->ed files [Andrew Wilson]
E-mail: Andrew.Wilson@cm.cf.ac.uk

it's probable that B5 will be combined, on the first pass, with:

E7) Send Last-modified header for server-side-included docs if
group XBIT is set. [Rob Hartill, by way of Andrew Wilson]

but I'll have to chat some with Rob H about this. If this is the case then
I'll boss for E7 as well.

> Take a peek at http://www.hyperreal.com/httpd, it's a start. Hoepfully
> will evolve into something decent in the next few days.

I can't imagine it staying unchanged for very long ;)

> Right now the
> cgi's are on my machine (www.steam.com), but I will be moving them over
> after they are cleaned up and debuged a bit more.
>
> My plans are to enhance the subimisson form with a bunch of
> instructions. Also the result will include instructions on
> where to ftp your patch code.

Well, I guess I'm happy doing all this by e-mail right now. A WWW summary
of e-mail traffic / patch info, would be a nice start, but if you want to
write some cgi to automate submissions then that's fine. I don't want to wait
for the scripts to be in place though. Let's keep it rolling please.

A suggestion:

1) All submissions should be (where possible) in the form of UNIX
context diffs, eg:

diff -c3 f1.c f1.c.old > f1.diff

made against the 'current' version of the server, initially NCSA 1.3R
taken from our own FTP archive, not NCSA's, just incase NCSA change
something in their 'R' release without letting us know [that's the
sort of sly thing I do sometimes ;) ]

2) For each patch, one person only should be responsible for submission.
In this example, if Rob H has a better idea for the patch then he tells
me, and the pair of us refit the patch and then I resubmit it.

So can we just mail a patch to this list or do we gotta wait for the FTP site to
go live? Tell me, someone? ;)

> Cliff

Cheers,
Ay.
Re: My life in films... [ In reply to ]
The ftp site is up, I've created a mode 777 incoming directory
in ftp://ftp.hyperreal.com/httpd for people to send in the patches.
have at it. Please either use my or rst's numbering scheme (I'll
sort it out later) when sending stuff somewhere in the file name.

I don't want anyone to wait to start integrating code. Let
the process catch up to what is being done. Keep 'em comming.
I still have patches to add from rst's list...should be done
in an hour or so.

Also who can ftp the 1.3R reference code we will be using?

Cliff
Re: My life in films... [ In reply to ]
>
> Well, I had a couple of minor patches to add, which were just
> omissions on my part from the earlier list (Bugfix: raise the
> queue-length argument to listen() to something reasonable; Bugfix:

didn't McCool or someone following up to one of his mailings say
that this needs to be a kernel fix in some (sunos?) systems.

> treat scripts returning Status: 30x the same as those returning any
> other explicit status code, rather than invoking the magic CGI
> handling for Location). However, steam.com doesn't seem to be up
> right now, so I can't even see Cliff's list, let alone modify it.

I have a simple patch for that. It also works the same for scripts
just sending a Location - all the script output is sent.

rob h
Re: My life in films... [ In reply to ]
Date: Thu, 9 Mar 95 11:05:53 GMT
From: Andrew Wilson <andrew@www.elsevier.co.uk>

Well, I guess I'm happy doing all this by e-mail right now. A WWW
summary of e-mail traffic / patch info, would be a nice start, but
if you want to write some cgi to automate submissions then that's
fine. I don't want to wait for the scripts to be in place though.
Let's keep it rolling please.

Well, I had a couple of minor patches to add, which were just
omissions on my part from the earlier list (Bugfix: raise the
queue-length argument to listen() to something reasonable; Bugfix:
treat scripts returning Status: 30x the same as those returning any
other explicit status code, rather than invoking the magic CGI
handling for Location). However, steam.com doesn't seem to be up
right now, so I can't even see Cliff's list, let alone modify it.

NB, a version of my patch list which incorporates the above two items
is at http://www.ai.mit.edu/tools/new-httpd/patch-list.txt --- perhaps
we can just use that until Cliff's scripts are working reliably. (I
think I've been maintaining the list on a fairly timely basis).

rst
Re: My life in films... [ In reply to ]
From: Rob Hartill <hartill@ooo.lanl.gov>
Date: Thu, 9 Mar 95 9:21:46 MST

>
> Well, I had a couple of minor patches to add, which were just
> omissions on my part from the earlier list (Bugfix: raise the
> queue-length argument to listen() to something reasonable; Bugfix:

didn't McCool or someone following up to one of his mailings say
that this needs to be a kernel fix in some (sunos?) systems.

The actual queue length is the *minimum* of two numbers: the queue
length requested by the daemon, and a limit imposed by the kernel. In
SunOS, and some other really stale BSD derivatives, that limit is
apparently compiled into the kernel, so the change to the daemon is
ineffective barring kernel changes. (If anyone has an easier way to
change it, even adb on /dev/kmem, please let me know).

However, even on those systems, the change to the daemon is benign,
and it is necessary in order to take advantage of more sensibly
configured kernels, be they SunOS or anything else.

rst

PS --- I do have some experience with this; my SunOS site was experiencing
frequent queue overloads until we raised the queue length parm to 128
in both the server and the kernel.
Re: My life in films... [ In reply to ]
On Mar 9, 10:50am, Robert S. Thau wrote:
}
} Well, I had a couple of minor patches to add, which were just
} omissions on my part from the earlier list (Bugfix: raise the
} queue-length argument to listen() to something reasonable; Bugfix:
} treat scripts returning Status: 30x the same as those returning any
} other explicit status code, rather than invoking the magic CGI
} handling for Location). However, steam.com doesn't seem to be up
} right now, so I can't even see Cliff's list, let alone modify it.

!@#$% power failure this morning...down for about 5 hours. It's back
up now (and everyone can add stuff). I placed no restrictions on
entering data.

I'd like to be sure that everyone who plan to do alot of hacking and/or
want's to help maintain the list gets an account on hyperreal.com from
Brian....send him that email :-) I see edits and stuff in email to rst's
list and I want to make sure everyone keeps in sync.

Cliff
Re: My life in films... [ In reply to ]
in response to the problems with the scripts...

I have a bit more work to do...I'm guessing it will take about
another two days to get the scripts in line. I'll be changing the format
a bit...and adding an "update" mode.

Thanks for the tips, and find one more escaped character I did not catch.

Cliff
Re: My life in films... [ In reply to ]
From: cliffs@steam.com (Cliff Skolnick)
Date: Thu, 9 Mar 1995 13:43:12 PST

!@#$% power failure this morning...down for about 5 hours. It's back
up now (and everyone can add stuff). I placed no restrictions on
entering data.

Cliff

Just submitted the two listed in my original note, and the one pulled
off the Net by Rob Hartill. First notes:

On the positive side, it's nice to have a brief bug summary. However,
it doesn't group patches by category. New bug reports go to the end
of the list; they really ought to be integrated with the other bug
reports (so those can be seen as a group, rather than interspersed
with long-term wish-list requests for functional extensions).

At least one input field is handled improperly --- I said I had code
for the first two (it's trivial in one case; I've got Rob Hartill's
fix for the other); it didn't believe me.

There's no obvious way to modify the writeup on an existing problem.
This comes up a lot --- today on my copy of the list I modified two
(added drtr's B1C, and changed the writeup on my P2), not to mention
the basic case where someone just wants to say, "I now have code for
this" (possibly in response to someone else's problem report).

Finally, a lot of the functional extensions that I had listed don't
seem to be there yet.

All this will come with time, I'm sure --- I just hope that with
Cliff's time evidently such a scarce resource, maintenance of this Web
structure doesn't take time away from the more vital business of
actually putting a server together.

rst
Re: My life in films... [ In reply to ]
One more note --- it needs to learn more about unescaping in
responses. The '%3D' in my writeup for patch 19 was typed as '='.

rst