The Big Changes (on both sides)

[Dirk Mueller]

On Mit, 08 Jan 2003, David Hyatt wrote:

> (1) The removal of KONQBLOCK.  I didn't quite finish this.  I am done 
> now.  I can post a diff if you'd like (a diff against our tree?) to see 
> it.

That would be very good (diff against CVS HEAD, always). 

I agree it was a terrible hack, and its great to 
see it gone. I'm curious on where it broke currently though. Haven't had 
problems with it according to user feedback the last few months. Also it 
appears to me that WinIE renders content that was moved to front as if it 
would be in a box with the same with constrains as the table itself. Mozilla 
doesn't do that, but I don't know if that matters. 

> (2) White-space changes and min/max width changes.  There are still a 
> few issues and regressions from this change.

Of course. again I'm very happy to see the white space handling where it 
should be. we also have collected some testcases regarding whitespace, did 
you see them?

> (3) Blocks inside inlines.  There are still a few issues here, in 
> particular with selection and event handling.

Also positioning should still be broken there. The selection stuff is going 
for a rewrite anyway. 

BTW, Lars was working a few days ago on improving the CSS parser, to be less 
bloated, probably faster and have proper support for CSS escape sequences. 
Just in case you have that on your todo..

> Changes that I think are doing pretty well at this point include the 
> layer changes, the FOUC changes and the margin collapsing code in 
> layoutBlockChildren.

I'm surprised by the FOUC change btw. especially the isInline "hack" doesn't 
appear clean to me. Also there might be races when external stylesheets 
load slowly. 

> Lars, can you comment on the state of your new table code?  I'm 
> interested in merging that back into our tree, and I'd love to get an 
> overview of where you are with that (and maybe a little background on 
> how the new code works).

I believe its a lot better than the old code. As more and more sites switch 
to table-layout: fixed it will give you a nice speedup on those pages as 
well. 

Topics I'm interested in include: 

- Don't allow subsequent <title> elements after the first to change the
title.

wow, other browsers do that ? But they still allow to modify it via 
Javascript, right?

- Reworked the list-style-position quirk for orphaned <li>s to not use
CSS.

why?

- Allow <form>s at all levels of a table, but demote them to leaves,
and make their subcontent into siblings of the <form>.

Very nice, this is what I was working on the last few days. Is this 
resembling the slightly weird behaviour Mozilla is exposing ?

- Removed the blockBidi hack from the parser.

How did you compensate the speed loss ?

- Removed the automatic flattening of "XML-style" tags. It is incorrect
to treat <foo/> as if it closed in HTML and results in errors in
rendering on lots of Web sites. New behavior matches Gecko and WinIE.

I think gecko does it when its a Xhtml site, and so do we AFAIK. WinIE 
doesn't thats known. 

do you have some more details on which constructs where used on those 
webpages, or which webpages they were ?

We've received so many bugreports from people who expected that flattening 
is working. 

- List items have been reworked to no longer make bullets float. All
bullets are now simply normal inline content. Added support for the
WinIE quirk that prevents nested list bullets from being on the same
line as ancestor lists (in strict mode this can be done). Matches
Gecko/WinIE.

Very nice. 

- Fixed several bugs with vertical-align.

Testcase / site ?

- Implemented size-adjusted and probability aware LRU in the loader
(LRU-SP) for improved handling of eviction from the loader cache.

Many thanks. was the old code very broken? I never really analyzed its 
behaviour. 


-- 
Dirk (received 279 mails today)