hc-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Aurelien Pernoud" <apern...@sopragroup.com>
Subject RE: Still problems with HttpRecoverableException and HTTPS
Date Tue, 15 Apr 2003 08:52:31 GMT

Jeffrey Dever a écrit :
>
> This could be serious if we are talking about a lost transaction.  Say
> you write a credit card payment successfully but can't read for some
> reason.  HttpMethodBase 1.43 would submit it again up to 100 times
> without programmer intervention or knowledge.  The read exception is
> pushed up to the programmer, so they can determine what to do with a
> failed write.
>
> Jandalf.

Although I understand the idea of separating request/write blocks, I must
say I strongly disagree with the fact that this should be fixed on the
client side.
When writing web applications, no one can tell wether or not everything will
happen fine. The connection can get lost at any time, the client can close
its browser after a request, whatever "bugs" that are not server side, but
related to TCP or client.
I don't know how browsers like mozilla/ie work when meeting such trouble.

Your idea of letting the dev choose what to do is great, but with current
CVS it's awful to do a retry :

-> execute mark method as used, so you can't simply retry the request, you
have to clone the method

With the oldest httpclient, the used was set to true but not retested in the
block, so everything went fine.

I've been stuck with this bug for a looooooong time and noone seems able to
tell what's happening there (For me the bug occurs under tomcat 4.1.x,
"sometimes"). Don't minsunderstand me, I know you guys are working on it, I
appreciate the great work already done in investigating, it's a very weird
bug.

But right now I'll simply apply the proposed patch that makes httpclient
work the way it used too, cause I'm facing a production time, and I already
used my old httpclient in tests here without any troubles reported.

Aurelien


Mime
View raw message