HTML Tidy has been upgraded on our servers, as of 04:40 UTC. We have had reports that some wikitext relies on fixed tidy bugs, and is thus broken, e.g. http://tinyurl.com/2orw8n . You can safely assume that we won't be simulating any Tidy bugs in MediaWiki to compensate, so these changes are here for good. If anyone can identify clearly broken behaviour in the new Tidy, then we can arrange to have it fixed.
-- Tim Starling
On 4/6/07, Tim Starling tstarling@wikimedia.org wrote:
HTML Tidy has been upgraded on our servers, as of 04:40 UTC. We have had reports that some wikitext relies on fixed tidy bugs, and is thus broken, e.g. http://tinyurl.com/2orw8n . You can safely assume that we won't be simulating any Tidy bugs in MediaWiki to compensate, so these changes are here for good. If anyone can identify clearly broken behaviour in the new Tidy, then we can arrange to have it fixed.
A point that seems to be biting a lot of people is the improper use of <span/> to surround block elements. This isn't proper.. so tidy fixes it.. but when tidy fixes it on a numbered list you get an odd result like
1. an item 3. another item
This can be solved by using div to surround block things like lists.
Here is an example: http://en.wikipedia.org/wiki/User:Gmaxwell/test
Gregory Maxwell wrote:
On 4/6/07, Tim Starling tstarling@wikimedia.org wrote:
HTML Tidy has been upgraded on our servers, as of 04:40 UTC. We have had reports that some wikitext relies on fixed tidy bugs, and is thus broken, e.g. http://tinyurl.com/2orw8n . You can safely assume that we won't be simulating any Tidy bugs in MediaWiki to compensate, so these changes are here for good. If anyone can identify clearly broken behaviour in the new Tidy, then we can arrange to have it fixed.
A point that seems to be biting a lot of people is the improper use of <span/> to surround block elements. This isn't proper.. so tidy fixes it.. but when tidy fixes it on a numbered list you get an odd result like
- an item
- another item
This can be solved by using div to surround block things like lists.
A lot of this needs to be detected by MediaWiki up front when the pages are saved rather than waiting for tidy to clean it up later. I deal with this breakage every time I import and test a dump file. MediaWiki could avoid a lot of these issues by detecting garbage when the pages are saved.
Jeff
On 4/6/07, Jeff V. Merkey jmerkey@wolfmountaingroup.com wrote:
A lot of this needs to be detected by MediaWiki up front when the pages are saved rather than waiting for tidy to clean it up later. I deal with this breakage every time I import and test a dump file. MediaWiki could avoid a lot of these issues by detecting garbage when the pages are saved.
This is a philosophical issue about how Wiki works. The lack of hard syntax rules that people must conform to facilitates discover and greatly softens the learning curve. ... it also means that our content will perpetually be rubbish syntactically.
I used to find this very distasteful. But a lot more time thinking about it, I don't feel as bad about it. What is bad is that our syntax is not well defined (and is, in fact, partially defined by tidy of all things) ... this means that making changes while remaining bug to bug compatible is nearly impossible.
I don't think syntax checking is a desired feature, ... but having a grammar which is well defined enough that changes can be made without breaking dependencies on unexpected aspects of our behavior would be very desirable. ... but doing so without breaking compatibility with the existing syntax is probably impossible.
On 06/04/07, Gregory Maxwell gmaxwell@gmail.com wrote:
I don't think syntax checking is a desired feature, ... but having a grammar which is well defined enough that changes can be made without breaking dependencies on unexpected aspects of our behavior would be very desirable. ... but doing so without breaking compatibility with the existing syntax is probably impossible.
It should be definable slowly if we work out just what the breakages will be and set bots loose to fix them.
Or just break them, *then* let bots loose to fix them.
- d.
On 4/6/07, Gregory Maxwell gmaxwell@gmail.com wrote:
This is a philosophical issue about how Wiki works. The lack of hard syntax rules that people must conform to facilitates discover and greatly softens the learning curve. ... it also means that our content will perpetually be rubbish syntactically.
It could be made sane on save, though, rather than on render.
On 06/04/07, Simetrical Simetrical+wikilist@gmail.com wrote:
On 4/6/07, Gregory Maxwell gmaxwell@gmail.com wrote:
This is a philosophical issue about how Wiki works. The lack of hard syntax rules that people must conform to facilitates discover and greatly softens the learning curve. ... it also means that our content will perpetually be rubbish syntactically.
It could be made sane on save, though, rather than on render.
Hmm ... could give unexpected results. "That's not the text I entered!" And bug reports.
OTOH, MediaWiki reporting it fixed syntax problems might be enough.
OTOOH, correcting a typo and getting back that 20 other changes have been made could be ... doing the unexpected.
- d.
On 4/6/07, David Gerard dgerard@gmail.com wrote:
Hmm ... could give unexpected results. "That's not the text I entered!" And bug reports.
We get that today.. except now it's the form of old versions mysteriously changing even though they can't be edited. With no obvious record of the changes.. and no easy way to confirm you weren't imagining what you saw the day before.
OTOH, MediaWiki reporting it fixed syntax problems might be enough.
We could also go to a hybrid approach once we get revision tagging: include a syntax checker (which could also check for other more style related issue), and then only allow people to set certain revision tags on revisions that pass the test... i.e. the articles overall syntax might be crud, but the stable version will always check okay. This will give a nice series of milestone versions which are not only good to human eyes, but also stand a chance of staying that way as the machinery changes.
OTOOH, correcting a typo and getting back that 20 other changes have been made could be ... doing the unexpected.
Right.. There lies the rub.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Moin,
On Friday 06 April 2007 18:26:31 Simetrical wrote:
On 4/6/07, Gregory Maxwell gmaxwell@gmail.com wrote:
This is a philosophical issue about how Wiki works. The lack of hard syntax rules that people must conform to facilitates discover and greatly softens the learning curve. ... it also means that our content will perpetually be rubbish syntactically.
It could be made sane on save, though, rather than on render.
Could be with a big red warning shown on preview:
"The version you are trying to save has been tidied up to prevent invalid HTML syntax. Please check ..."
It is a bit like my POD extension actually tells you before you save (on preview) that your POD is malformed.
All the best,
Tels
- -- Signed on Fri Apr 6 20:31:01 2007 with key 0x93B84C15. View my photo gallery: http://bloodgate.com/photos PGP key on http://bloodgate.com/tels.asc or per email.
"Time flies like an arrow; fruit flies like a banana."
-- Groucho Marx
Simply an alert of "Note that you have syntax errors" (without automatic machine "fixing") would be helpful.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Platonides wrote:
Simply an alert of "Note that you have syntax errors" (without automatic machine "fixing") would be helpful.
That would be extremely user-unfriendly. "Wikipedia won't let me ediiiiiiit articles! Something about bad sintax!!!!!!11oneone"
On 07/04/07, Edward Z. Yang edwardzyang@thewritingpot.com wrote:
That would be extremely user-unfriendly. "Wikipedia won't let me ediiiiiiit articles! Something about bad sintax!!!!!!11oneone"
NOBODY EXPECTS THE SYNTAX INQUISITION!
Rob Church
Sorry . but that is funny
DSig David Tod Sigafoos | SANMAR Corporation
-----Original Message----- From: wikitech-l-bounces@lists.wikimedia.org [mailto:wikitech-l-bounces@lists.wikimedia.org] On Behalf Of Rob Church Sent: Friday, April 06, 2007 17:03 To: Wikimedia developers Subject: Re: [Wikitech-l] Tidy upgraded
On 07/04/07, Edward Z. Yang edwardzyang@thewritingpot.com wrote:
That would be extremely user-unfriendly. "Wikipedia won't let me ediiiiiiit articles! Something about bad sintax!!!!!!11oneone"
NOBODY EXPECTS THE SYNTAX INQUISITION!
Edward Z. Yang wrote:
Platonides wrote:
Simply an alert of "Note that you have syntax errors" (without automatic machine "fixing") would be helpful.
That would be extremely user-unfriendly. "Wikipedia won't let me ediiiiiiit articles! Something about bad sintax!!!!!!11oneone"
You could still save it. Take it as an information of "something is wrong", just like Firefox console error or not passing the w3c validation, not just checking if it renders on quirks mode. Some templates are *hard* to check for unmatched {{ }} or tags. This would give you a hint about it. You can keep it hidden by default if you prefer, but as i originally said, having it would be helpful.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Platonides wrote:
You could still save it. Take it as an information of "something is wrong", just like Firefox console error or not passing the w3c validation, not just checking if it renders on quirks mode. Some templates are *hard* to check for unmatched {{ }} or tags. This would give you a hint about it. You can keep it hidden by default if you prefer, but as i originally said, having it would be helpful.
Perhaps. But it kind of defeats the purpose, as the ones who know enough to actually run the code against a validator would also be the ones least likely to produce non-standards compliant code (I'm not talking about unmatched braces: just silly things like divs inside of font tags).
And of course, we must do all we can to stop the growth of overly complex templates.
"Simetrical" Simetrical+wikilist@gmail.com wrote in message news:7c2a12e20704061126i13131ad4n550e1ae6afc55865@mail.gmail.com...
On 4/6/07, Gregory Maxwell
gmaxwell@gmail.com wrote:
This is a philosophical issue about how Wiki works. The lack of hard syntax rules that people must conform to facilitates discover and greatly softens the learning curve. ... it also means that our content will perpetually be rubbish syntactically.
It could be made sane on save, though, rather than on render.
Surely most of the objections to this could be solved by adding an extra edit into the history.
(cur) (last) 11:39, 9 April 2007 MediaWiki Syntax Corrector (help) (cur) (last) 11:39, 9 April 2007 HappyDog (Talk | contribs) (fixed typo)
- Mark Clements (HappyDog)
On 09/04/07, Mark Clements gmane@kennel17.co.uk wrote:
Surely most of the objections to this could be solved by adding an extra edit into the history. (cur) (last) 11:39, 9 April 2007 MediaWiki Syntax Corrector (help) (cur) (last) 11:39, 9 April 2007 HappyDog (Talk | contribs) (fixed typo)
Nice one!
- d.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Tim Starling wrote:
HTML Tidy has been upgraded on our servers, as of 04:40 UTC. We have had reports that some wikitext relies on fixed tidy bugs, and is thus broken, e.g. http://tinyurl.com/2orw8n . You can safely assume that we won't be simulating any Tidy bugs in MediaWiki to compensate, so these changes are here for good. If anyone can identify clearly broken behaviour in the new Tidy, then we can arrange to have it fixed.
TinyURL is broken, use: http://tinyurl.com/2xzn72
On a more relevant note, any code relying on Tidy bugs must be shot and refactored. I wholeheartedly agree. :-)
wikitech-l@lists.wikimedia.org