[Bug] Re: SSL3_WRITE_PENDING:bad write retry

[Ethan Schoonover]

On Wed, Feb 2, 2011 at 02:14, Sebastian Spaeth <Sebastian at sspaeth.de> wrote:

> I believe the right thing would be to catch the correct exception and
> the right place, and to close down the connection properly and retry
> with a new SSL connection.
> I still haven't looked deeply into how the thread pool works, so I am
> not sure what we should be doing. One thing that will help is if we can
> see at which places we get exactly which exceptions.

I'm out of my depth here, but the first thing I think of is TCP
congestion control. In addition to the connection shutdown/reconnect,
is this also a case for some sort of TCP-like "slow start" which
increases based on SSL connections failures? Thus, in cases where, for
example, 3 clients are contending for 12 connections where only 10 are
allowed by the server, a simple TCP like slow start congestion control
mechanism would solve the problem without having clients (or at least
well behaved clients) continuously pile back onto the server for a
connection?

I may be totally off base here, or even proposing something best
handled at a different layer. I'll bow out of this thread and let
wiser heads carry on :)

Ethan