From: Jorge A. C. <jc...@ne...> - 2001-09-23 09:38:14
|
Hi everyone! Now that 0.6.1 has achieved a significant stability gain, we can focus on several other tasks that were waiting in the priorities queue. Note that the stabilization is not completed, but advanced enough to let other tasks to be started. ----------------- Attribute parsing ----------------- As I posted before (Subject: "Attribute parsing"), this is a high priority task that's pending. I received emails from Bruno and J=F6rgen stating that they were willing to work on it, but not knowing exactly were to start. My current schedule contemplates working on this, the networking part of plugins, https and auth. So If I start with the latter I'll have little time to help you both, so J=F6rgen: if you feel confident enough to work on attribute parsing, starting from my previous post, please let me know so I can focus on networking; if not, let me know also so I can start here. Note that there're other interesting tasks in this post that if you feel more comfortable with, can be as helpful to work on. -------------------------------------- Bug 216 (answers without content/type) -------------------------------------- Yes, I also had this problem with ebay (made a little hack, and won my bid!). Since that moment the idea of handling this case the rigth way is present. Note that the problem arises from a http answer lacking the content/type info; the RFC says it SHOULD be present, so it's not an error :( The solution is simple, do you remember magic numbers and the 'file' command? Well, that's the way to go. I don't mean calling 'file' and parsing its output, but to examine the magic numbers file (and its format) for the small set of MIME types dillo supports: image/{png, jpeg, gif} text/{plain, html} and afterward, to create a custom function that returns the MIME type of a file, by examining a few bytes from the start of it. Ex: gchar *a_Mime_get_content_type(const gchar *FewBytesString); After having this function, it's just a matter of binding it to header parsing. -------------------- Dicache memory usage -------------------- The cache system (cache and dicache) uses a lot of memory for images, because it stores the original format and the RGB decompressed one. Flushing the dicache is not easy because image widgets use the dicache buffers directly. The right solution is to implement a memory usage threshold on the dicache, but that would require a significative effort... As a workaround, there's a posibility of binding the 'about:splash' method (because it doesn't use images) to a dicache flushing function. i.e. whenever the splash screen is requested, the handling method asserts there's only a single browser window open and flushes the dicache (after displaying the splash). This is very far from a clean solution, but may serve those of you that feel in a dire need of it. ------- Plugins ------- One of the most procrastinated topics in dillo :( After attribute parsing is done, I'll be choosing between plugins and the networking part of https, cookies and auth. ------------------------ https, cookies and auth. ------------------------ There were three issues to this: 1) Stability (almost achieved) 2) The specific implementation layer and network support for them. 3) https pages usually require working cookies. As Tor-Ake and J=F6rgen have advanced with 2 and 3, I'll need to know what this schemes require from the networking layers (for instance, cookies require sending back stored cookies). If you guys work together, and specify what you require from the networking layers, and keep on improving the specific implementation, it will come a time when I'll be free to implement the underlaying support and hopefully then, it'll become a matter of merging. Note that cookies should provide dillorc options for: - rejecting all cookies - accepting only if the RootUrl and server match. - accepting all - [confirm] -- this must be very well thought, for not annoying the user. ----------------- Frames workaround ----------------- Before the final implementation of frames, a workaround (as dicussed on the list) would be a nice addition. I'll be waiting until Livio finishes testing and polishing the current prototype. ------- History ------- We are working with Eric on it. Expect a full creative solution to be there sometime :) I'm crossing fingers. ----------------------- BUG 205 (visited links) ----------------------- This is not a bug, but a feature! In dillo, visited link sematic is "cached". (it has proven useful) Note: 205 will be removed soon from BugTrack. ---------------- nowrap and width ---------------- The workaround for this construct was removed in concordance with dillo's html parsing policy ([Project Notes]). Also note that the HTML 4.01 spec says WIDTH is a "recommended" cell width (hints in dillo's implementation), and NOWRAP "disables automatic text wrapping for this cell", and it adds "NOTE: if used carelessly, this attribute may result in excessively wide cells". So current implementation is compliant with both. Regards Jorge.- |