Adrian, Oleg thanks very much for your responses! I don't have much time to respond right now in detail so I thought I'd select one aspect and write a combined response.

Kalnichevski, Oleg wrote:
[ ... ] If you wish I can itemize
>>exactly what is wrong with these libraries and tell you what I need.
>
> I believe it would be time well spent, and your input will be highly
> appreciated.

Ok, here goes. Btw, unless specified otherwise, "HttpClient" means "HttpClient Library"

- One of the main defiencies of the innovation.ch library in my use of it is the threading issue: it creates a thread whenever it tries to establish a connection to a server. However, if you have, for instance, a transient network failure then multiple instances of these threads may build up quickly and you then run out (there is a mode where it can share connections but this is buggy too).
So: requirement 1 is that HttpClient minimizes creation of Threads. From what I've seen so far, this is already pretty much satisfied.


- The problem that innovation.ch solved here was timeouts. I'm sure you all know that the Sun implementation (as of 1.3) doesn't support setting of them.
So: requirement 2 is that HttpClient supports timeouts on response reading. Again, satisfied.


- Redirects. I need to follow redirects, but only up to a limit. This limit is actually not just based on the number of redirects in a chain but also on other factors, which are not derivable based on anything HttpClient knows about.
So: requirement 3 is that I need to arbitrarily stop following redirects at any given point in the chain.


- Redirects again. They must not be followed immediately. This is related to the previous point, but I think different. I need to be able to recieve a redirect and then say "Oh, I'll deal with that later, if ever". Adrian: this is one thing that I think supplying a "retry handler" would *not* give me.
So: requirement 4 is that I should not be forced into `evaluating' the redirect immediately.


That's it for now, I think. I don't think I'm wanting *too* much on the features front. I'm more in need of robustness in the face of all those errant web servers out there :-)

I've been holding off using HttpClient until it stabilised and requirement 1 has been kicking me up the ass as recently as today so it's likely I'll pick up the latest stable release very soon (next few weeks) and start seeing what else I can criticise, but this time in more detail :-)

--
Mike


--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]



Reply via email to