| |
| |
| |
|
Page: 1 2 3 4
Comments:
<0> edtrager: I know about hanzi principles (I know japanese), but not so much about autohinting
<1> especially when you get entries that map to more or less the same results
<1> ie, if the text is "sans" and the user changes the font to "dejavu", there won't be any change
<2> autohinting has a set of `rules' which are applied to glyphs based on Unicode ranges
<2> (this is autohinting in FreeType)
<0> edtrager: however, bitmap fonts are loathed at least in Japanese, and are generally banned from fonts.conf
<2> for example, rules for CJK fonts have been added
<0> wl: yeah, but what kind of rules are these?
<2> rules for Arabic are missing
<1> edtrager: is there any effort towards making grayscale bitmap fonts for hanzi?
<3> fonts should be browseable by script, language support or cl***ification
<3> that would be more useful than aliases
<1> (and if every user changes the browser font to "sans", then why is there a "serif" alias at all? I ask half-seriously)
<2> mathrick: unfortunately, those rules are very complicated; for example, the CJK stuff is a 42k C source code file
<4> Raph: I have not investigated the details -- Wen Quan Yi has both Chinese and English web pages : http://wqy.sourceforge.net/cgi-bin/enindex.cgi
<2> mathrick: part of the `autofit' module in Freetype
<2> mathrick: those rules are of heuristic nature; there is always room to improve them
<5> moyogo: I agree, that's why the smart grouping from in the font widget proposal sounds very good: http://www.unifont.org/fontdialog/ what do you think?
<4> Moyogo: I really would like to find the time to code at least most of the non-GUI functionality that I proposed in http://www.unifont.org/fontdialog/.
<6> yosh: I think what would be a killer would be to add an icon before each font name
<1> edtrager: the samples shown are bilevel only
<6> trees are nice but users want flat-level
<4> Raph: What do you mean by "bilevel" only? You mean only two levels of nesting of categories?
<0> nim-nim: what would this icon do?
<5> nim-nim: there's so much useful metadata we could flag up from the binary fields in the fonts
<6> and that means putting the info about a font (hinted or not, suitable for printing) newt to it
<1> grayscale bitmaps are a good technical way to compromise between sharpness and smoothness, but they are poorly supported; i think the _real_ reason for this is that nobody in the proprietary world ever did grayscales, possibly due to lack of copyrightability
<3> yosch: it would be the right place to start :)
<7> yosch, one of the big problems with drop down lists is they are effective with machines with (say) 50 font families on, but not for machines with 500 font families on.
<1> ed: bilevel = black and white only, no gray
<6> yosh : at the simplest level ***ociate an icon to each cartegory in your cl***ifier and just put the icon next to fonts in a flat list
<2> BTW, FreeType supports grayscale bitmaps
<6> you can keep the tree view in the font manager
<6> for the font selector in OO.o for example, it's way overkill
<4> Raph: OK, sorry I misunderstood. So Wen QuanYi are Black and White only. Well, that is certainly easier to do ... I think those bitmaps are manually-tuned.
<3> are we still on topic?
<5> yep, this is where the fontmanager would allow you to create sets to make the fontlist more manageable
<3> ok... :) I got lost
<8> moyogo: we got sidetracked :)
<1> wl: that is good news indeed
<1> perhaps that information is not as widely known as should be
<3> so would that "fontmanager" be at a different layer than fontconfig?
<1> if people _are_ going to pour development effort into bitmap fonts, then perhaps they should be aiming for the more ambitious goal of quality graymaps
<6> How about : we use a MRU in a simple 20-entries drop-down with icons which show font categories, with another widget (+) lanching the big font manager?
<4> And only on topic #2 : In order to discuss which fonts are best for which scripts, I first would like to ask what is the state of including the Graphite Module in Pango now : Is this a fait accompli so the next release of Pango ?
<0> edtrager: behdad said he sees no problem with reviewing the patch soon, that was at GUADEC
<0> dunno the status right now
<0> one (minor) requirement was to drop the libpangographite
<8> edtrager: I have seen demos of Graphite (ping yosch), it looks good. I suppose we need more widespread testing and evaluation.
<5> both behdad and gl***eyes will be at the desktopcon in Ottawa. Pretty sure they'll work it out there
<9> it's pretty worked out, I just need to get the work done
<4> I will certainly try to provide my opinions on how to fill in the chart at http://wiki.freedesktop.org/wiki/Software_2fFonts for which fonts are best for which scripts. I will ***ume that Graphite support will be coming soon enough so that, for example, Padauk can be the Unmodulated (i.e. "Sans") default for Myanmar.
<6> BTW, sans and serif ****, modulated/unmodulated are versy user-unfriendly, how about simple and complex?
<9> afaiu at least the pango-query-modules part of the patch will be in the next unstable release, and the rest should be as well, as long as I get it done in time
<9> (pangographite that is)
<7> I think there are probably two font interfaces needed. A drop-down, and some sort of "finder" type interface that copes with families.
<8> gl***eyes: that's great. I suppose it will mean that adding the graphite module, one can easily get Graphite in their existing system.
<5> just a little patience then folks :-D
<8> :)
<4> Note that the Prahita (http://sourceforge.net/projects/prahita) project has released a Pango module for Myanmar which works with the author's Masterpiece Uni Sans font : I hope this Prahita Myanmar module also makes it very soon into the Pango release.
<9> simosx: yes, and it will be up to the distros whether they include it in their packages or not
<8> gl***eyes: thanks
<8> I think at this point we can go on to agenda items 3 and 4, dealing with the fonts.conf file.
<8> Currently, fonts.conf does not offer a flexible way to deal with competing fonts;
<10> greetings, is there any work ongoing on http://www.unifont.org/fontdialog/ ?
<8> Some background on the fonts.conf file: among others, it specifies the aliases with a bias towards western languages.
<4> Prokoudine: A few minutes ago I mentioned that I, having written the http://www.unifont.org/fontdialog/ stuff, have thought a lot about writing some of the non-GUI parts of the functionality in C++. But my problem is not having enough time and having too many projects. I don't know of anyone else working on it ...
<11> The idea of aliases is that it provides a fallback mechanism for when a glyph is not found in one font. It can then drop down the list and find the glyph in another font.
<8> fonts.conf also specifies "preference" lists for each alias; glyphs come from the fonts in the preference list; if the top font has glyphs missing, fontconfig searches the second in the list, and so on.
<8> there is also the mechanish of "orthographies"; if a font has partial support for a language, it is ignored in this type of matching.
<11> One issue with this mechanism is that if two different fonts contain the same two seperate languages, and you want to use font 1 for one language and font 2 for the other, you will be unable to do so with fontconfig as it stands.
<6> but fontconfig also ***umes fonts are nicely limited to single scripts, when in practice they cover as many scripts as they can to avoid recreating common glyphs
<8> for this to work well, one needs to make sure the "orthography" files are up to date for a language; but not too strict so that it rejects all fonts.
<5> simosx: what might be useful is a little GUI to handle the order of that prefer list
<1> jhobson: it sounds like the functionality you desire is to select not a single named font, but a list of preferred fonts in priority order
<12> lo
<6> yosh: you can check easily that fontconfig does not allow expressing some natural things now
<8> The "preference" selection takes place when you select the alias, currently "sans", "serif", "monospace".
<11> Actually, what I need is a mechanism to allow me to place fonts in a font config alias and have it pick my prefered font even when a font above it supports that language.
<6> like use A B C fonts for greeks and B A C for cyrillic
<11> nim-nim: exactly
<5> nim-nim: spot on
<8> yosch: I would think that the preference list would be set on a low level; that the user would not have to worry about it.
<6> simosx: in fact the first step is make the default settings **** less, user-level GUI is the second one
<11> On the agenda I submitted a patch to fontconfig that allows fontconfig to hide language out of fonts. It does not go to the level of unicode ranges, but I think it is easier to understand for end users.
<8> the current mechanism for font preferences is to use the "language" field in fonts.conf.
<4> jhobson and nim-nim: I'm trying to understand what you want: You want to be able to specify distinct font preference lists for each script and/or maybe even each language?
<11> simosx: Yes, but that mechanism will find the first match, which is not necessarily the best match for that font, given that it may be there to support a different language
<13> lo MrB :)
<8> nim-nim, yosch: yeah, we have big issues with the default settings. the user-level gui is desirable.
<12> hi davelab6
<2> sorry, I have to leave--mother nature is asking for sleep
<2> happy talking!
<5> keithp mentionned at the GUADEC that he was looking at per block filtering
<8> jhobson: I suppose you mention what I said about orthographies above. Indeed, that's a side mechanism, not the main one.
<14> what nim-nim is proposing would also be important for Chinese
<6> edtrager: actually, what I personnaly want is for fontconfig to manipulate blocks-of glyphs whithin fonts instead of full fonts
<6> jhobson's proposal does maybe 80% of that
<8> yosch: indeed, keithp said he would be happy to receive a patch that clearly blocks a section of glyphs from a designated font (defined in fonts.conf), so that we do not have to create forks of fonts to get the same functionality.
<6> which is ok than 0% now
<8> to the best of my knowledge, noone is working yet on the functionality that keithp suggested.
<15> yosch: not per-block filtering, but the ability to mark sections (by unicode) of fonts as if they didn't exist at all
<6> simosx: keithp's suggestion is good for blacklisting terminally ill glyphs
<6> not sorting somewhat-ok glyphs
<14> except with chinese the problem is often style - traditional and simplified character sets overlap but often have a different style.
<15> nim-nim: yeah, the other question is whether we can do per-language 'not so good' measurements
<15> victory747: that's handled by language tags already
<11> keithp: actually it is not, we have fonts with Japanese characters that are used in Chinese text output. It really iritates the Chinese.
<16> Unicode is awkward that way
<15> jhobson: then your chinese fonts are broken
<14> it's even worse in a text file where you have no meta-information
<15> mjg59: that's why we tag fonts for language support
<6> keithp: I'm sure we can collect enough user feedback so we can sort fonts per language manually
<16> keithp: But an arbitrary unicode file doesn't contain a flagged language
<16> It's not just Japanese/Chinese where this is a problem
<15> mjg59: so we fall back to locale in that case
<11> The Japanese font claims to support Chinese, and has a subset of those characters
<16> keithp: So how do you handle a mixed chinese/japanese document?
<6> the problem is how to teach fontconfig the local preferences once we know them
<14> yeah, the order for substitution needs to change based on locale, but even that is not perfect
<16> It's not actually just chinese/japanese
<1> mjg: in that case, you clearly need language tags in the doc
<16> You get somewhat similar problems with gaelic
<15> mjg59: you either have language tags for doc sections, or you pick one. Any better plans?
<1> and devanagari - the rules for which ligs you select vary from language to language
<6> raph: the word processor needs to now the language anyway to do spell checking correctly
<16> keithp: I'd have preferred Unicode to have tackled this properly
<16> I'm not blaming fontconfig :)
<15> mjg59: so would the Japanese :-)
<1> mjg59: a little late for that :)
<16> AIUI, there are language metadata codepoints in Unicode, but they're deprecated?
<1> mjg59: yes
<16> Yeah. I think that's somewhat short sighted
<11> Providing a unicode range blocking mechanism for fontconfig would be great, but other than people on this IRC, I don't know anyone that could work in that environment effectively.
<1> the philosophy is that you have a markup layer that specifies the language tag as well as all other presentation-layer stuff, like bolding and so on
<16> It gets a bit into "Damnit, I know what this specification is for" at the expense of actually being useful
<6> I think that for gui we just use the user locale info, for browser the language tags, for office suite the word processor info
<8> jhobson: the blocking mechanish is for distributions to provide an adequate default for their users.
<6> ****s for plain text users though
<1> mjg59: it breaks stuff like cut-n-paste; it's not hard to see why they chose what they did
<12> simosx: ***uming the fonts the user has has the glyphs in the right areas..
<16> raph: Hm. Interesting point.
<11> If it is only for distributions to handle, then it kinda ****s to allow users to add new fonts.
<6> jhobson: well, allowing them to add fonts nevertheless is better than forbidding them to add fonts
<0> raph: lang codepoints or lack thereof breaks c&p?
<6> and font distributors can learn to provide fontconfig info to fump in ~/fonbts
<11> Exactly, users should be able to add new fonts, but they also need a mechanism to get around the problems we are looking at.
<1> i think there are two ease-of-use issues that are worth considering - _installing_ a new font, and configuring your defaults so that your new font is selected in aliases such as "sans"
<1> mathrick: lang codepoints. sorry for being unclear
Return to
#freedesktop
or
Go to some related
logs:
#perl
java-config Malformed CLASSPATH
mod_python ZipImportError
dmraid rebuild ICH7R
mysql_connect() [function.mysql-connect]: Access denied for user 'root'@'localho
ldap +libpam +libnss-ldap +etch
#debian
#gaim
_Raptor_ ircnet
pacpl + rpm
|
|