Re: [PATCH] HTTP_request

Hi! Richard Heyes wrote:

1) A patch I sent to Richard a lo-o-ong time ago to make appending [] to variables optional (like the one that was applied to Net_URL)

Can't remember for sure, but I beleive I rejected this.

I don't remember either. But it was a two-part patch for Net_URL and HTTP_Request having the identical functionality (making braces [] after the array variable optional). The Net_URL part was applied and released, the HTTP_Request part was not. Thus braces after array values can be optional in GET requests but are mandatory in POST ones now. :]

3) Stefan Walk's patch to implement Observer pattern and to do a lot of

other

stuff.

Not seen this.

It is not exactly a patch, but a rewritten HTTP_Request. I have a file around and may put it for review somewhere. Background: Stefan proposed a class that could download files via HTTP and display a progress bar. The proposal was rejected and he was told that it is better implemented as an add-on to HTTP_Request. He communicated with Christian Stocker about this and created a modified version of HTTP_Request where he changed a lot of other unrelated things as well. After some time the discussion stopped and the proposed changes were never commited. I think it is possible to extract only the Observer-related parts and to create a manageable patch out of this.

I can work on integrting the patches and preparing a new major release of HTTP_Request, as its current state limits the possibilities of development

for

my own HTTP_Client.

As long as the code is good quality, and previously rejected patches aren't applied, then this shouldn't be a problem. Also, I wouldn't want to see HTTP_Request becoming bloated due to requirements in HTTP_Client, if you know what I mean.

Yes, I understand your concerns. The only thing that is somewhat related to HTTP_Client is Observers, as Client already has these and it will be better for the users to make them compatible.

There is also a problem with redirection support in HTTP_Request. My

personal

POW is that redirection does not belong to the package at all and its

addition

was a BC break done in a minor 1.1.1 release.

Too late, it's in there now.

There are numerous problems with current implementation. To properly support redirects you'll have to copy a *large* chunk of code from HTTP_Client, including the whole HTTP_Client_CookieManager class. And the *main* problem is that it is on by default. I suggest at least making it off by default. And the last question, on procedure: can I commit to HTTP_Request or should I create patches and send them to you?