Talk:Comparison of layout engines (CSS)

From Wikipedia, the free encyclopedia

This article was nominated for deletion on 24 February 2007. The result of the discussion was keep.

Contents 1 Archive of old discussions 2 Criteria of CSS versions? 3 Suggestion: Last checked versions 4 What about Prince? 5 Test cases 6 I clarified some issues on the CSS2 CSS2.1 and CSS3 aural and speech properties. 7 CSS 2.1 and CSS3 8 Experimental as a yes (green) 9 CSS3 and Browser specific extensions 10 Zero width or height 11 Links to properties 12 Proprietry properties 13 Add Mozilla to list of engines 14 Add HDi 15 Add WinWord version of Trident 16 Soon to be no longer Presto?? 17 Presto has full css 2.1 'support' 18 Order of properties 19 Presto cursor issue?

[edit] Archive of old discussions

• All discussions not added to since 1 jan 2007: Talk:Comparison_of_layout_engines_(CSS)/Archive 1--itpastorn (talk) 12:36, 20 February 2008 (UTC)

[edit] Criteria of CSS versions?

REC CSS2? CR CSS2.1? latest CSS2.1? or WD CSS3? I see there're some inconsistencies in this article. Let's say the white-space property, it seems we're now using CR CSS2.1 as the criteria due to "Partial" support of Gecko, but the Webcore and Presto should also be "Partial" if we really do so. In this case, Presto should be "Partial" in counter-* property. And content property support of Gecko should be "Yes" again while that of Presto is somewhat between CR CSS2.1 and WD CSS3 level. Another issue is the changes from CR CSS2.1 to latest CSS2.1. Since latest CSS2.1 is not in public yet, it's hard to find out all the changes until now. But its section 12.4 has been rewritten as far as I know. If we use this as criteria, that row will become "Partial" in KHTML column as what KHTML implements is the specs in CR CSS2.1. It's also crucial to define how can we put "Yes" into the table? "Yes" if the layout engine has partial support? buggy support? or only what they claims they support?

N.B. Some "experimental" support listed here are actually the superset of the formal CSS specs. How should we comment on this? Jiu 19:08, 25 Apr 2005 (UTC)

That's the problem we face in the current table layout. Suggestion? You can put your suggested layout here. P.S. "Yes" is put for full support only (based on CSS2.1). --minghong 13:27, 25 Apr 2005 (UTC)

By the way, please try to offer footnotes for the "partial" support. Currently, for some of the items we don't know why it is "partial" but not "full". I know that's difficult, and some are actually made by me... --minghong 18:47, 25 Apr 2005 (UTC)

Someone removed the references to CSS1. Can anyone explain why ? As far as I know, CSS1 is still a standard. Is it just because the comparison is not interesting anymore, since every engines compared support it ? What about the history (engines version n#) ? I think it doesn't cost much to keep that information. --Fenring 13:42, 29 July 2007 (UTC)

Fenring, all the information is still there. KorangaR just changed "CSS1" to "CSS2", which it has been all the time anyways, since some CSS1 properties get extended by CSS2 etc. The alternative would be to copy all properties that get extended to the respective row and mark support for CSS1, CSS2 different for the same properties. I don't think this is helpful in the case of CSS1. We could try something like that for CSS 2 / CSS 2.1, though. Anyways, I think it's a good idea... maybe we need some text explaining why there's no "CSS1 properties" on the page? --Grey 16:23, 29 July 2007 (UTC)

Thanks Grey. I understand and agree with the removing of CSS2.0. As 2.1 corrects a few errors in 2.0 specs, and thus was intended to replace 2.0 instead of extend it. Though, I think it is helpful to know which properties belong to which level of the official specifications (1, 2.1 and 3). I would rather keep the CSS1 properties in CSS1, and add a note explaining CSS2 fully includes CSS1 spec. (And another note explaining CSS3 fully includes CSS2.1, by the way). For extended properties, we could write in the cells, if it is the 1, 2.1 or 3 version of them which are supported. What about this kind of table layout that visually explains that (I don't like it though : I'd rather use a note) ?

CSS version
3	2.1	1	important	Weight increasing
			/*Comment*/	Comments
			@import	Import stylesheet
			@charset	Character set
			@media	Media-specific rules
			@page	For paged media
			@namespace	Namespace declaration
			@font-face	Define font

I don't like this, either. You needn't do this for each table anyways, text stating the situation would be sufficient, since you'd still have to mark which version of the property is supported. The idea to put the supported version in the cells seems clear enough, I just wonder how it'd look like. --Grey 11:14, 30 July 2007 (UTC)

Ok. So, let's lose the information of what property belong to CSS1. I can't imagine a situation where it would be useful, anyway. But I think at least the CSS1 row in the General overview table should stay. Say I am building a CSS1 website. I'd like wikipedia to tell me what browsers' versions to advice on the homepage. --Fenring 14:28, 30 July 2007 (UTC)

You're right, of course, didn't see that. It should stay in the overview no matter what. --Grey 18:53, 31 July 2007 (UTC)

I disagree. There is no such thing as an CSS 1 site, since you may use the subset of CSS that was in version 1, but the newer spec may differ, and in such instances the old spec is obsolete. When two specs disagree, the newer one always obsoletes the older one - and it is according to the new spec browsers must comply. Therefore: CSS 2.1 is the only spec browsers must follow today. CSS 2.0 has been deprecated in its favor.[1]. Building a CSS 1 site therefore means using that subset of CSS 1, that has not been altered at all in 2.1. Good luck with that!--itpastorn (talk) 12:46, 20 February 2008 (UTC)

You mean a CSS1 conforming document is not rendered the same by a CSS1 conformant UA and a CSS2.1 UA ? Interesting. I didn't know that. I was aware of the corrections between 2.0 and 2.1 only. What are the CSS1 properties that disagree with the 2.1's ones ? Do you think we can integrate that information into the article ? Because even is CSS1 is obsoleted, it still exists. I may want to use all features of CSS1 if I target only CSS1 UAs. CSS is not a web-only spec. --Fenring (talk) 14:27, 20 February 2008 (UTC)

References for changes between CSS 1.0 to CSS 2.0 and CSS 2.0 to CSS 2.1. The only difference between CSS 1.0 and CSS 2.1 that seems to be documented are differences in tokenization. --Execvator (talk) 16:28, 20 February 2008 (UTC)

[edit] Suggestion: Last checked versions

It would help maintaining it. A no always means that it will probably be included in a future version. So for anybody who whants to help keeping this huge list up to date it would be a relief to know what version was the last checked. E.g. in Safari it is written "Property XYZ: No" - I got no idea when with what version this was checked back the last time.

I don't get your question. A "no" just means a "no": It isn't supported in the past, not now. Future? We don't know. It doesn't mean anything else. --minghong 01:51, 19 August 2005 (UTC)

I get it. He means that there are always updates (especially Safari and Firefox) which fix things and implement new (css) features, so maybe the version of Safari which was used to check the css property should be noted. It'd make maintaining the page easier. 85.1.125.112 20:37, 31 January 2007 (UTC)

I think that would clutter the page up big time and make the DOM article in the series unmaintainable. We don't need to be superprecise. As far as Gecko and Presto go, they'll be maintained good enough anyway (and IE... *laugh*). Then again, we can't used it for Safari/WebKit as some people started adding dates to features/properties with "No template" indicating Nightly support. So the layout engines that could possibly benefit are KHTML and iCab... and this does (imho) not justify the mess -- Grey 01:50, 1 February 2007 (UTC)

[edit] What about Prince?

Since we speak about layout engines, not browsers ... why not include Prince XML formatter? It's the forerunner in CSS implementation as far as I know. Albeit only for print media ... Grey 23:29, 5 June 2006 (UTC)

Add it if you have data. 203.33.3.12 05:44, 17 August 2007 (UTC)

almost done ;)

[maybe somebody can help and add a few more data] Mabdul (talk) 13:13, 25 March 2008 (UTC)

[edit] Test cases

Where the test suites can be obtained to check / augment the tables? We should make these test suites publicly avaiable to make this information verifieble, to conform to Wikipedia standards. --Maxim Masiutin 00:13, 15 January 2007 (UTC)

[edit] I clarified some issues on the CSS2 CSS2.1 and CSS3 aural and speech properties.

I clarified some issues on the CSS2 CSS2.1 and CSS3 aural and speech properties. CSS2.1 is a draft and should not be cited as a current recommendation. Also true for CSS3. I think this article could do better to keep that issue in the forefront of the reader’s mind. Also, the section did not keep clear the difference between CSS3 speech and aural modules. If we’re going to include CSS3 discussion, those should be kept straight. Things like "volume" and 'voice-volume' are not the same property. In CSS3 one is an aural property and the other is a speech property. A similar distinction could be made between 'voice-balance' and 'azimuth' or 'balance'. --SoTuft 08:39, 22 February 2007 (UTC)

[edit] CSS 2.1 and CSS3

I think there's nothing wrong with the article discussing these working drafts. However, I think the article should make clear that these are proposed recommendations (working drafts, candidate recommendations, etc.). Saying that something is deprecated in CSS2.1 is misleading. It would be better if it said something like “CSS2.1 proposes the deprecation of …” Likewise for CSS3. It's definitely of interest since browsers are already anticipating CSS2.1 and CSS3 and so this article should probably include what browsers support the proposed features in each .

However, many of the additions of CSS2.1 are in CSS3, so even if CSS3 was never adopted it wouldn't really matter. Many of the properties and the like (at-rules, keywords and modules) that are deprecated in CSS2.1 are reintroduced in much the same form in CSS3. If certain CSS3 modules reach recommendation status before CSS2.1, it really becomes confusing. I almost think the deprecations of CSS2.1 are meant to be a marketing gimmick so that one implementation or another can claim “FULL CSS 2.1 support”. The only thing that is deprecated by CSS2.1 that remains so in CSS3 (that I can think of off the top of my head) is the 'display:marker' value and the 'marks' property from the CSS2 paged media module. There may be a few others, but they're part of nonpublic drafts and shouldn't be discussed here. --SoTuft 08:39, 22 February 2007 (UTC)

Just as another example of the ways it’s misleading to discuss CSS2.1 as a done deal look at the @font-face rule. Currently the article says that “@font-face removed in CSS 2.1” However, the CSS3 WebFonts Module (which is at the same “Last Call” status) includes the @font-face rule. So again, if they both reach recommendation status at the same time, then CSS2.1 merely serves as a marketing document so an implemntation can claim full CSS2.1 support. It doesn't really help web designers or web consumers since fully supporting CSS2 would be a more expressive implementation. It sounds better for a browser to say it fully supports CSS2.1 than to say it fully supports CSS2. However, CSS2 give designers, content creators and content consumers more to work with than CSS2.1. --SoTuft 08:39, 22 February 2007 (UTC)

Another thought on disentanbling the current CSS recommendations from proposed recommendations woudl be to add explicit separation between CSS2 and CSS2.1. Perhaps repeating properties when the meaning changes from one recommendation to another (it doesn't happen that much).. That way it would be eawier to show whether an engine implemented one or the other of the recommendations. Also an introductory table like this could help:

	Trident	Tasman	Gecko	WebCore	KHTML	Presto	iCab	Prince
CSS1	6.0	Yes	1.0	85	Yes	7.0	Yes	Yes
CSS2 (2.0)	Partial	Slight	Partial	Partial	Partial	Partial	Partial	Mostly
CSS2 (2.1)	Partial	Partial	Mostly	Mostly	Mostly	Mostly	Mostly	Mostly
CSS3	No	No	Slight	Slight	Slight	Slight	Slight	Slight

Since 2.1 deprecates Aural and the page context, I would imagine Gecko and probably Presto, WebCore and KHTML will be nearly 2.1 complete this year. However, I think it's disingenuine to give that much credit when the recommendation changed to match their feature set (and also for a recommendation that is not yet a recommendation). Then I think in the other tables adding ana dditional section for CSS2 and CSS2.1 (explicitly differentiating them), would help make things clearer. Any other thoughts?--SoTuft 08:49, 22 February 2007 (UTC)

[edit] Experimental as a yes (green)

I changed over the experimental designations to yes in the article. Let me explain here why. The experimental designation has become the norm throughout the UA development world. Any property or selector that is only part of a draft recommendation is only going to get implemented with the name-spaced prefix. However, content creators are free to start using these properties and selectors right away. They simply require repeating the declarations for each implementation. Once the recommendation is finalized, you can expect those browsers that already have the prefixed implementation will quickly offer the non-prefixed version. This is good practice for everyone since it let's early adopters try out the features and the implementations; identify bugs early; and ensure the final rollout has the greatest amount of interoperability. So I think these make the most sense as green in the table since that's the most we should expect from CSS3 properties and selectors. Once CSS3 becomes a recommendation, it might make sense to change these back to partial (though I don't think tat we'll have the window of opportunity to do that). --SoTuft 09:25, 22 February 2007 (UTC)

[edit] CSS3 and Browser specific extensions

I believe that the sections for which the W3 hasn't at least got to 'Candidate Recommendation' status should either be removed or have a thumping great warning attached (something akin to {{future}}) -- it's potentially unfair to judge a browser on a stard before it has been finalised (after all, that's why the Box model was "wrong" in IE, as it correctly implemented one of the WD implementations)

I also don't think that browser proprietary extensions should be listed at all (all those starting with a '-') -- Ratarsed 10:21, 25 February 2007 (UTC)

Concerning CSS3, the situation is difficult... we should include modules that have been CRs (Selectors, primarily), but the CR documents are in part obsoleted by their respective Working Drafts now. It's just an obscurity of the W3C's standards track that specs go back in status and we have to deal with it. It would be a shame to exclude the selectors module, just because it is not CR any more (implementations are forthcoming). Regarding "proprietary extensions"... they aren't listed unless they copy the behaviour of a spec'ed property, which makes them "experimental". See above section for opinions about those (I don't like that they are there/"green", but some disagree)... --Grey 19:34, 25 February 2007 (UTC)

I'm the one who suggested changing the experimental implementations of the working drafts to green (and I went ahead and did so). However, I would agree that a better appraoch would be to simply not include CSS 2.1 or CSS 3 at this point. I think the proposals should actually reach recommendation status before they're included in this article. Until they reach that stage they're in constant flux. It's difficult to even find realiable sources that assess whether the various rendering engines have completely, partially, implemented the feature and whether it's buggy or dangerous or such things. Also, by focussing on the stable recommendations, we could focus more attention on the subtle differences and similaritiesin these rendering engines. CSS 2.1 and CSS 3 open a whole other can of worms that complicates the issues. If CSS 2.1 reaches recommendation status it could be added back in. Also as far as CSS 3 is concerned, I think each of it’s modules should be treated separately by this article, just as the W3C is doing. --207.168.161.13 17:31, 1 March 2007 (UTC)

[edit] Zero width or height

Could you please cover the fact that IE6 doesn't accept "height: 0" or "width: 0" for an element. When trying to use it, the element will still be visible. Dunno what W3C says about it, but Mozilla supports it though. This feature is useful when putting an element that is initially invisible into a table-like container. Width is left as "auto", but height is set to 0. Thus there'll be no vertical gap to its contiguous elements, but once its visiblity is toggled, it won't possbily increase the width of its parent, as it would if you had used "display : none" => "display: block".

It's an overflow:visible bug which is fixed in IE7, i will notify it thank you. -- KorangaR 23:07, 28 July 2007 (UTC)

[edit] Links to properties

Adding links to the appropriate specifications for each element/etc (like http://www.w3.org/TR/REC-CSS2/selector.html#id-selectors for "id selectors") would be adventageous. 203.33.3.12 06:37, 17 August 2007 (UTC)

[edit] Proprietry properties

Would the addition of non-W3C attributes/elements etc be a good idea? If not on this page, then a separate one? 203.33.3.12 06:37, 17 August 2007 (UTC)

+102:01, 9 February 2008 (UTC) —Preceding unsigned comment added by 84.169.191.7 (talk)

Comparison of layout engines (Non-standard HTML) done! fell free and help ;) Mabdul (talk) 18:59, 7 April 2008 (UTC)

[edit] Add Mozilla to list of engines

Pre-Gecko browsers supported parts of CSS2, and probably have more users than iCab. Perhaps instead of adding a new column, the information could be included in the Gecko column, say M4 for Navigator 4. 203.33.3.12 02:35, 27 August 2007 (UTC)

ok but have anybody information/refs for this? Mabdul (talk) 12:30, 7 May 2008 (UTC)

[edit] Add HDi

HDi Interactive Format is the system used to create HD-DVD menus. It uses CSS. It's support for CSS elements should be included. 220.253.157.16 01:37, 1 September 2007 (UTC)

this should be no problem, but I have no references which tags/attributes are supported! Do you have any lists? Mabdul (talk) 13:22, 25 March 2008 (UTC)

[edit] Add WinWord version of Trident

WinWord uses a cut down version of Trident as it's rendering engine. It is shared by Outlook 2007. Details can be found here: http://msdn2.microsoft.com/en-us/library/aa338201.aspx. 203.33.3.2 03:18, 11 October 2007 (UTC)

maybe next month I will do this... The List seems to be complete ;) Mabdul (talk) 13:22, 25 March 2008 (UTC)

[edit] Soon to be no longer Presto??

As of Opera 9.5 and 10, opera is changing thier rendering engine (including the name) to Kestrel and then Peregrine respectively. Do we just keep it presto here and footnote it, or change Presto to just Opera as a general blanket term for the closed source engines it has always had. Chase-san 06:53, 29 October 2007 (UTC)

As far as I know, Opera isn't replacing its layout engine. Kestrel and Peregrine are just code names of the future releases.--Fenring 09:38, 29 October 2007 (UTC)

Presto is the name of the Opera rendering engine, has been since version 7.0. There are two branches of Presto so far - Core and Core-2. Merlin was the code name of the Opera 9.0 release, which used the Core branch of Presto. Kestrel is the code name of the to-be Opera 9.50 release, which uses the Core-2 branch of Presto. Peregrine is the code name of the to-be Opera 10 release, which will also use the Core-2 branch of Presto. -- liorean (talk) 10:48, 17 January 2008 (UTC)

Just to clarify, Core-2 is not the rendering engine, but the platform (which includes the rendering engine). [2] --Grey (talk) 17:59, 17 January 2008 (UTC)

[edit] Presto has full css 2.1 'support'

This section details the support for css, not thier correctness, if that was the intent, wouldn't the section be named not "CSS version support" but "CSS version completeness", or something of simular value? This is my reasoning behind Presto's yes in the css 2.1 box, this may also be grounds for firefox or simularly supporting browser to also be noted as having full 'support' of the css. If this is not the intended, (as people get religious about certain browsers appearing better then the one they use), then I suggest we change the name of the category. --Chase-san (talk) 21:48, 2 January 2008 (UTC)

As you can see, I have reverted the recent edits that made false claims in Presto's favor, so I don't think that's necessary anyway. Indeed, though, it is not "correctness" but "full implementation" that counts. There simply is no one single implementation that is "all correct". I have a problem with "version completeness", though. It just doesn't seem to make sense as a compound word. "Support" is often defined the way we use the terminology here. So I don't see any reason to change it in the future, unless common usage changes. --Grey (talk) 14:49, 3 January 2008 (UTC)

However, if you would get a version of Opera (now 9.25) and test visibility: collapse, it is supported, and it works correctly, if you test white-space: pre-line, it is also supported and works correctly (which actually I think a link was included on that one). Though I admit I did accidently omit the table columns visibility collapse comment. Just because they we're added only to Presto does not make the information wrong, but if you would like I can go and retest everything listed with all the browers my system can support. (Trident, Gecko, Presto, and perhaps Webcore if it is in windows versions of Safari) --Chase-san (talk) 00:32, 5 January 2008 (UTC)

Ummm here is a test-page for the white-space property: http://www.csdgn.org/test/white-space.html --Chase-san (talk) 01:02, 5 January 2008 (UTC)

First, I just tested and don't see Opera 9.25 supporting pre-line nor collapse. But that's not the problem : you have to find a reliable source that says it has tested these values. I can't find any that says these are supported. All the sources (even opera itself) say Opera doesn't support them. --Fenring (talk) 11:03, 5 January 2008 (UTC)

Strange, it supports them here... *checks again* and now it doesn't. Okay, I give you that, that is seriously odd as I swear it was working, but cannot contend with the results of this test, which show they do not. I apologize for all the problems I caused then. --Chase-san (talk) 12:01, 6 January 2008 (UTC)

[edit] Order of properties

Are the property categories (Box model, border, text etc) in any specific order? --Stefán Örvarr Sigmundsson (talk) 00:36, 19 March 2008 (UTC)

[edit] Presto cursor issue?

I've noticed that in Presto when putting the cursor property in a :hover or :active selector it will only show the intended cursor on the first time hovered, but not on subsequent times, don't know what you guys would want to list that as. —Preceding unsigned comment added by 64.199.245.2 (talk) 15:05, 2 June 2008 (UTC)