httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marc Slemko <ma...@znep.com>
Subject Re: several messages
Date Sat, 01 Feb 1997 19:58:59 GMT
On Sat, 1 Feb 1997, Randy Terbush wrote:

> 
> FTR - I'm not convinced we have found the solution here either.
> My sky has been falling in some other areas and I have not had
> much time to offer constructive data or suggestions. One fact
> remains that we are calling shutdown on closed sockets. Dunno
> if this is a symptom or just a fact of life.

A small number of them are to be expected because there is always going to
be some place where the client can shut it down before you.  That's life.
However, there may be more to it.

On Sat, 1 Feb 1997, Ben Laurie wrote:

> No, I am concerned, but I'm not convinced that there is _still_ a problem. You
> think lingering_close() does it - but isn't it the case that a small number of
> OSes _need_ lingering_close() to fix other problems?

Two issues.  One the broken SVR4 stuff that flushes the buffers when a
close() is done on a socket without sending the data.  This is fixed by
lingering_close() and was the reason presented on the list for needing it.
If we did not need lingering_close() for other reasons, the workaround of
doing a shutdown() then a close() should work.

> 
> If it is being used when it doesn't need to be, then that is a problem.
> 
> OK, so I've woken up now. I've just checked, and it would seem that _only_
> A/UX disables lingering_close(). But I thought our decision last time round
> was to only enable it when absolutely necessary? What happened?
> 
> That is, we decided that the default would be to not do a lingering close,
> and to enable it for those OSes that needed it, in Configure. That clearly
> hasn't happened. Why not?

Because Roy brought up the "real" reason for lingering_close, as discussed
by him with rst and others.  This reason makes sense and tests have shown
that there _are_ cases where clients have problems with PUTs and
persistent connections without lingering_close().

On Sat, 1 Feb 1997, Jim Jagielski wrote:

> Ok, ok.... I'll be quiet...
> 
> It's kinda obvious that no one else seems as concerned about those
> OSs without FIN_WAIT_2 timeouts. I seem to be bugging people too
> much as well...

I am concerned.  Give me the magic words to fix it, I'll jump with joy.
If I had a dump of the packet flow between a client and a server which
left the server with a connection in FIN_WAIT_2, and not just due to the
known bug with keepalive timeouts, I could begin to guess at the problem.


Mime
View raw message