hc-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Roland Weber <http-as...@dubioso.net>
Subject Re: HttpAsync Design - please review
Date Thu, 01 Jun 2006 13:42:04 GMT
Hi Oleg,

> It's great. I wish I had a half of your ability and patience to document
> design alternatives I confront. 

Thanks a lot. Although I have to admit it has little to do with patience.
The problem just got so big that I couldn't keep the thoughts in my head
anymore :-) I know it will save time in the end, so I do it now.

> (1) "Applications must process responses as they arrive"
> 
> If the problem cannot be resolved at the API level, an effort should be
> made to provide a detection mechanism throwing an IllegalStateException
> if an attempt is made to process responses in a wrong sequence. 

We could define a maximum timespan in which a handle has to be closed
after the response is available. For blocking IO, there will be a thread
waiting for the connection to become available again, so that thread
can enforce the timeout and force-close the handle.

> (2) red vs cyan

As mentioned in reply to Mike's mail, I'll pick this discussion up
next week.

> (for instance, AsyncHttpProcessor _may_ no longer be required)

For pipelining, the preprocessing/sending part has to be called
by a different thread than the receiving/postprocessing part.
HttpRequestExecutor just doesn't offer the interface to do that.
I have one more idea to get rid of AsyncHttpProcessor with less
intrusive changes to http-core than previous attempts. I'll tackle
that when I'm back to coding in July or August.

cheers,
  Roland

---------------------------------------------------------------------
To unsubscribe, e-mail: httpclient-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: httpclient-dev-help@jakarta.apache.org


Mime
View raw message