On Mon, 2007-01-29 at 16:44 +0100, Roland Weber wrote: > Hi Oleg, > > >> 2. attribute in the HttpClient > > > > I tend to prefer 2. I am strongly not in favor of storing interfaces in > > HttpParams. CredentialsProvier was my biggest design blunder in the > > HttpClient 3.x days. > > I was beginning to wonder where to draw the line between params and > context. That's a good distinction. Parameters are for constants only, > meaning instances of final classes such as String, Integer, HttpVersion. > Except that HttpVersion is not final. So rather non-functional classes, > which only _represent_ a value and where the actual instance doesn't > matter. The stuff which would be a "valuetype" in IDL. > > > Why would you want a different connection re-use strategy on a per > > request basis in the first place? > > Good point. The only reason would be to prevent re-use of the connection > in specific situations. If the situation is known in advance, it is > better to send a "Connection: close" header with the request. If it is > determined dynamically, some yet to be defined "abort" method can be > used. No need at all for overriding the re-use strategy. > > >> Should we have a NoConnectionReuseStrategy implementation > >> of the interface in HttpCore? > > This question remains open. >
Sorry. Missed this one. Yes, I think we should Oleg > thanks, > Roland > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
