The bugs turned out to be not of my making. There are too many spots in the HttpMethodBase class and its derivatives where statusLine is blindly assumed to be always non-null. Cheers Oleg
-----Original Message----- From: Kalnichevski, Oleg Sent: Freitag, 31. Januar 2003 17:03 To: Commons HttpClient Project Subject: RE: [PATCH] PostMethod & PutMethod revision (take 2) Jandalf Got ya. I have just managed to reproduce the behaviour of "buggy" servers that ignore Expect: 100-continue header and discovered BIG FAT NASTY bugs in my code. So, I am currently busy busting those bugs. I'll proceed implementing your suggestions once all bugs are dealt with. I'll look into possibility of simplifying request buffering as well Stay tuned. Good thing come to those who wait Oleg -----Original Message----- From: Jeffrey Dever [mailto:[EMAIL PROTECTED]] Sent: Freitag, 31. Januar 2003 16:54 To: Commons HttpClient Project Subject: Re: [PATCH] PostMethod & PutMethod revision (take 2) > > >>Internally we should just have one "body" datamember, which would be an >>InputStream. When setting the request body, if the streamBody() form is >>used, its trivial. When the stringBody() is used, create a >>StringInputStream which is easy. When the parametersBody() form is >>used, url encode it to a string, and use that as the input stream, also >>easy. >> >> >I am not sure if we can totally do away with additional byte[] buffer member >variable, unless you are prepared to completely sacrifice ability to repeat requests >(if case of re-authentication, for instance). It may be a reasonable thing to do, >though, as it turns out Post & Put requests MUST NOT be automatically redirected >without user intervention according to RFC 2616 > > No redirection does work in our favour. Not sure what to do about the authentication issue. It would be really lame to post a multi-megabyte file only to be challenged for authentication and have to resend. We can buffer small bodies, but not big ones. With stringBody and and parametersBody at least we know how big they are, and perhaps there could be some decision to buffer. The client really should do a HEAD before a POST so authentication is performed before large files are sent over the wire. Jandalf. --------------------------------------------------------------------- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]