Serious concerns about font rendering regression
Hi, since 10.0, I noticed a regression with the font rendering quality, especially the kerning is degraded, compared to 9.3. The situation persists with 10.1 (or even got worser). I already filed a report against 10.0: https://bugzilla.novell.com/show_bug.cgi?id=133086 only to realize, that I'm soliloquizing. What I've learned from trying to fix the problem myself is that fontconfig and possibly qt3 uses an unofficial API from treetype2 to render stuff, and had a hard time to adjust freetype2 cvs to provide the hooks. Unfortunately, I suffered from some crashes using it (could be unrelated), but not enough time to examine the reasons, thus reverted my local changes that time. For what is worth it, I'm using the "Bitstream Vera Sans" truetype font for all sans serif uses. Getting back on topic, I hadn't compared 10.0 with 10.1, but 9.3 with 10.1, and besides a much worser kerning, bold glyphs are notedly less detailed (zugelaufen). Does nobody noticed this regression and is concerned, too? It could be, that I'm picky in this respect, but font rendering quality is a very important factor in user experience and acceptance. Pete
Am Samstag, 4. Februar 2006 23:08 schrieb Hans-Peter Jansen:
Hi,
since 10.0, I noticed a regression with the font rendering quality, especially the kerning is degraded, compared to 9.3. The situation persists with 10.1 (or even got worser).
I already filed a report against 10.0: https://bugzilla.novell.com/show_bug.cgi?id=133086 only to realize, that I'm soliloquizing.
What I've learned from trying to fix the problem myself is that fontconfig and possibly qt3 uses an unofficial API from treetype2 to render stuff, and had a hard time to adjust freetype2 cvs to provide the hooks. Unfortunately, I suffered from some crashes using it (could be unrelated), but not enough time to examine the reasons, thus reverted my local changes that time.
For what is worth it, I'm using the "Bitstream Vera Sans" truetype font for all sans serif uses.
Getting back on topic, I hadn't compared 10.0 with 10.1, but 9.3 with 10.1, and besides a much worser kerning, bold glyphs are notedly less detailed (zugelaufen).
Does nobody noticed this regression and is concerned, too?
It could be, that I'm picky in this respect, but font rendering quality is a very important factor in user experience and acceptance.
How should I interpret the yelling silence on this matter? Sure, people are busy with other things, but I wonder, that there's no interest in the quality of visual appearance in our favorite distro? The matter is, I can avoid not (anymore) supported hardware but I cannot ignore font render regressions, especially since 9.3 did better than successors (not, that it was perfect either). BTW, in the light of the whole free vs. protected discussion, isn't enabling TT_CONFIG_OPTION_BYTECODE_INTERPRETER via freetype2-bc.patch also relevant here? I really don't care the license issues involved here, but in my own tests, the font render quality was better with the BC interpreter disabled (as it is the case in 9.3). OTOH, freetype2 is also a strong moving target, and victim of one of the dirtiest patches, too, by exporting an internal interface (ft2-stream-compat.diff). Sure, other projects are involved here too like fontconfig and qt, but this problem isn't that new either. I would like to know more on the rationals behind these decisions, but nobody from the SUSE want to speak up. All I harvested is some bashing on the ft2 ML, how deeply they dislike this procedure, but just bashing me, a suffering user, won't change anything on this topic.. Pete
On Sat, Feb 18, 2006 at 12:52:48PM +0100, Hans-Peter Jansen wrote:
How should I interpret the yelling silence on this matter? Sure, people are busy with other things, but I wonder, that there's no interest in the quality of visual appearance in our favorite distro?
I looked just at the two screenshots provided in the bugreport and I don't see any difference between the two. Perhaps you need to describe better what the differences is that one needs to look at. I am not saying that there is no difference. I am saying that I don't see it and that may well be the case for other people as well. houghi -- Nutze die zeit. Sie ist das Kostbarste was wir haben, denn es ist unwiederbringliche Lebenszeit. Leben ist aber mehr als Wert und Arbeit, und das Sein wichtiger als das tun. Johannes Müller-Elmau
Am Samstag, 18. Februar 2006 13:19 schrieb houghi:
I looked just at the two screenshots provided in the bugreport and I don't see any difference between the two. Perhaps you need to describe better what the differences is that one needs to look at.
The kerning is very different. Look at the distance before and after the first "e" of "Bearbeiten", also the "s" of "Lesezeichen". Both appear outstanding unlike those from 9.3. Here's an * where I think, distances are to big: 9.3: B*earbei*ten A*ns*i*c*ht Gehe z*u Les*ez*eic*hen 10.0: B*e*arbe*ite*n An*s*i*cht G*eh*e zu L*e*s*eze*i*ch*e*n But maybe, it's just me, and everybody else is fine with it. I'm not a typographer, but it must hurt in their eyes like it does in mine..
I am not saying that there is no difference. I am saying that I don't see it and that may well be the case for other people as well.
I've my test and work machine stand side by side, and with switching back and force, the difference is pretty obvious. Pete "das Auge isst mit" Jansen
Hans-Peter Jansen schrieb:
Am Samstag, 18. Februar 2006 13:19 schrieb houghi:
I looked just at the two screenshots provided in the bugreport and I don't see any difference between the two. Perhaps you need to describe better what the differences is that one needs to look at.
The kerning is very different. Look at the distance before and after the first "e" of "Bearbeiten", also the "s" of "Lesezeichen". Both appear outstanding unlike those from 9.3.
Here's an * where I think, distances are to big: 9.3: B*earbei*ten A*ns*i*c*ht Gehe z*u Les*ez*eic*hen 10.0: B*e*arbe*ite*n An*s*i*cht G*eh*e zu L*e*s*eze*i*ch*e*n
But maybe, it's just me, and everybody else is fine with it. I'm not a typographer, but it must hurt in their eyes like it does in mine..
This is funny. While I agree that the rendering in 10.0 is more inconsistent, I strongly disagree that the distances are too large. I would want the large distances between all characters. But then again, I'm the only one who thinks it is important to be able to distinguish between rn and m. Regards, Carl-Daniel -- http://www.hailfinger.org/
Am Samstag, 18. Februar 2006 15:14 schrieb Hans-Peter Jansen:
Am Samstag, 18. Februar 2006 13:19 schrieb houghi:
I looked just at the two screenshots provided in the bugreport and I don't see any difference between the two. Perhaps you need to describe better what the differences is that one needs to look at.
The kerning is very different. Look at the distance before and after the first "e" of "Bearbeiten", also the "s" of "Lesezeichen". Both appear outstanding unlike those from 9.3.
Here's an * where I think, distances are to big: 9.3: B*earbei*ten A*ns*i*c*ht Gehe z*u Les*ez*eic*hen 10.0: B*e*arbe*ite*n An*s*i*cht G*eh*e zu L*e*s*eze*i*ch*e*n
But maybe, it's just me, and everybody else is fine with it. I'm not a typographer, but it must hurt in their eyes like it does in mine..
I am not saying that there is no difference. I am saying that I don't see it and that may well be the case for other people as well.
I've my test and work machine stand side by side, and with switching back and force, the difference is pretty obvious.
Pete "das Auge isst mit" Jansen
They actually look like different fonts, the rendering on the "B" is distictly clearer and the difference in the size of the two loops is much more pronounced. In the second one there is much more anti-aliasing and "blurring", saying that the "ea" (as an example) in the second image is much clearer, and the kerning is better. The little "b" in the first image is better, the n's and A's are about the same, the "G" from Gehe is a little clearer in the first one, but again the "ehe" is better from teh second image... You are correct, the Kerning is off, but not in a way that would worry me in day-to-day use, the "blurring" or the characters is more of a problem to me, as "out of focus" things (not just text) make my left eye go funny very quickly. I am very fussy when it comes to setting up my machines - and one of the reasons I never used 8.n as a desktop machine, after 20 minutes or so, I'd have to take a break of several hours before I could work again. In that respect the 9.n and 10.n have been a huge improvement - I think the move to xorg was the turning point in my eyes (sorry, about the pun). My 10.1 machine is down at the moment, waiting for the 10.1 B4 update. I haven't had much time to experiment so far. I'll look more closely at it once I have B4 installed (I started messing around with B3, but due to the significant changes in B4 and a heavy schedule, I didn't give B3 that much attention... Dave
On Sat, Feb 18, 2006 at 03:14:42PM +0100, Hans-Peter Jansen wrote:
I've my test and work machine stand side by side, and with switching back and force, the difference is pretty obvious.
I have saved them and going back and forth between the two files. As they are not the identical size, seeing the real differences is a bit hard. Also the font is acually too small to really nitice the difference, unless you indeed look at first one and then the other. To me it seems that the lowercase E is a bit different, but that is all I could see even after enlarging it. Perhaps the font is updated. Again with such a small font it is very hard to tell. Best try to use a mch larger font and perhaps idential dimensions for the pictures to show much more clearly what the difference is. Other the the difference in the E (perhaps, could be due to the low quality picture) I realy see absolutely no problem with it. houghi -- Nutze die zeit. Sie ist das Kostbarste was wir haben, denn es ist unwiederbringliche Lebenszeit. Leben ist aber mehr als Wert und Arbeit, und das Sein wichtiger als das tun. Johannes Müller-Elmau
Op zaterdag 18 februari 2006 16:07, schreef houghi:
I've my test and work machine stand side by side, and with switching back and force, the difference is pretty obvious.
I have saved them and going back and forth between the two files. As they are not the identical size, seeing the real differences is a bit hard. Also the font is acually too small to really nitice the difference, unless you indeed look at first one and then the other.
To me it seems that the lowercase E is a bit different, but that is all I could see even after enlarging it. Perhaps the font is updated. Again with such a small font it is very hard to tell. Best try to use a mch larger font and perhaps idential dimensions for the pictures to show much more clearly what the difference is.
Other the the difference in the E (perhaps, could be due to the low quality picture) I realy see absolutely no problem with it.
You should open them in konqueror side by side, using: konqueror -> window menu (left next to help) -> split window vertically. This shows the attachment next to eachother precisely. Making a good comparison possible. -- Richard Bos Without a home the journey is endless
On Sat, Feb 18, 2006 at 08:24:48PM +0100, Richard Bos wrote:
You should open them in konqueror side by side, using: konqueror -> window menu (left next to help) -> split window vertically. This shows the attachment next to eachother precisely. Making a good comparison possible.
Tried that and it makes the comparison more difficult then with swapping the two files very fast from the one to the other in gqview. houghi -- Nutze die zeit. Sie ist das Kostbarste was wir haben, denn es ist unwiederbringliche Lebenszeit. Leben ist aber mehr als Wert und Arbeit, und das Sein wichtiger als das tun. Johannes Müller-Elmau
Am Samstag, 18. Februar 2006 21:05 schrieb houghi:
On Sat, Feb 18, 2006 at 08:24:48PM +0100, Richard Bos wrote:
You should open them in konqueror side by side, using: konqueror -> window menu (left next to help) -> split window vertically. This shows the attachment next to eachother precisely. Making a good comparison possible.
Tried that and it makes the comparison more difficult then with swapping the two files very fast from the one to the other in gqview.
What I do in such a case is opening the link with the konqueror context menu option "open with..." and manually entering kuickshow (or the like). Then positioning both windows vertically. Pushing some "+" keys in both windows (magnify) let the rendering differences appear more obvious. Since both are the same font, one could have the impression, they're different, but it documents the huge differences of freetype 2.1.9 and 2.1.10 font rendering. Now you can imagine, that simply using a bigger font doesn't make any sense at all, since it results in totally different results.. Pete
On Saturday, 18. February 2006 21:05, houghi wrote:
On Sat, Feb 18, 2006 at 08:24:48PM +0100, Richard Bos wrote:
You should open them in konqueror side by side, using: konqueror -> window menu (left next to help) -> split window vertically. This shows the attachment next to eachother precisely. Making a good comparison possible.
Tried that and it makes the comparison more difficult then with swapping the two files very fast from the one to the other in gqview.
Use something like that: http://ktown.kde.org/~coolo/regression/output/unsorted/RESOLVED-62962-2847.h... Greetings, Stephan
I have saved them and going back and forth between the two files. As they are not the identical size, seeing the real differences is a bit hard. Also the font is acually too small to really nitice the difference, unless you indeed look at first one and then the other.
No, if you make everything 20mm high, you never see any difference in rendering because there are so much more pixel per letter height to render the glyph. The issue always is how to render fonts readable at small sizes. I have to agree with HPJ, the rendering in https://bugzilla.novell.com/attachment.cgi?id=56884&action=view is downright crappy compared to https://bugzilla.novell.com/attachment.cgi?id=56885&action=view In the first, word parts to both sides of the "e" look disjointed, worse, the "e" looks like it's falling over towards the left, or in other words, is rotated CCW by 10° - see "B e arb". Yuck. Typographically, you want letters of the same word to be horizontally visually equidistant (ligatures here are 1 letter by themselves), because it improves readability. This has nothing to do with "rn" looking the same as "m" - that's failure to maintain minimum separation between letters. The most serious problem by far for me is the new rendering thing in the ghostscript 8.x x11 device, which is downright useless for low 8.x and not nearly acceptable for the current 8.x, compared with 7.07. http://bugs.ghostscript.com/show_bug.cgi?id=688496 https://bugzilla.novell.com/show_bug.cgi?id=140100 (Ouch, need to get back to that.) Volker -- Volker Kuhlmann is possibly list0570 with the domain in header http://volker.dnsalias.net/ Please do not CC list postings to me.
participants (7)
-
Carl-Daniel Hailfinger
-
David Wright
-
Hans-Peter Jansen
-
houghi
-
Richard Bos
-
Stephan Kulow
-
Volker Kuhlmann