I noticed today that a new version of CURLHandle (1.9) was released this week and Dan seems concerned that its reason for being may be eclipsed by WebKit. That seems unlikely to me, even in the longer term view of things.
I just can't see developers who have adopted CURLHandle as being likely to give it up just because a system framework supports most of their needs. This becomes more important if you want to support older systems (this seems less important to developers lately and yeah, that bugs me). By virtue of being wrapped around libcurl (I was surprised to see LDAP support mentioned — I wonder what that means?), there is widely deployed (and tested) functionality that doesn't seem to be available in WebKit along with access to lower levels of the protocol(s) (if needed, all source is there).
Among the many problems Windows developers have faced over the last few years when using the IE engine are boundary conditions (and whose code gets to own them) and updates. WebKit is no different. The engine's handling of cookies, headers, certificates, persistence and even protocols is still in flux and subject to significant changes. At some point (if it hasn't happened already) an API or behavioral change is going to break applications (and then the fingers will fly!) Worse (from purely a client perspective) is what could happen if support for WebKit wanes.
No real answers will be forthcoming from this paranoid corner of the universe, but I think I'd be inclined to use WebKit for display and stay away from the transport side of things.
Posted by Dave at April 3, 2004 07:17 PMAmen.
I really, really worry about the stability of WebKit's transport layer. A brief foray into the source code a few months ago laid bare a caching algorithm that was so primitive to be worthless. It did not recognize nor generate any sort of 'last-modified' header, but instead assumed that cached resources never go stale. (By comparison, Netscape 2.0 had proper 304 support!) A small, evil part of me wants to believe that this was done to create the marketable illusion of speed, but I suspect it is simply a programmer omission.
I'm still stuck in MacOS 10.2 so maybe WebKit has gotten better, but Apple seems to be enamored by its display functionality and is ignoring its transport features. So I don't blame anyone for using CURLHandle, not at all.
Posted by: Rob Menke on April 4, 2004 03:20 AM commLink