If you thought Windows rasterizing was complicated ...

Jens Kutilek's picture

... then check out the bug reports related to Linux rendering for the Ubuntu font family.

https://bugs.launchpad.net/ubuntu-font-family/+bugs

But beware, only true font nerds can find this entertaining on a Saturday afternoon ;)

clauses's picture

Oh that's nothing. A few days work, and that should be fixed. But I must say that having the bugs in a bug-tracker is a good idea. Versioning and bug-tracking is a good idea. Did anyone do a versioning system for UFO's yet? I know of one person who have done some work on such a system, but it's not done.

Khaled Hosny's picture

Most of the bugs are about incomplete glyph coverage, glyph design, kerning etc. few about rendering issues and it includes rendering issues on Windows too since the font is being tested through a wen app as a web font.

blank's picture

I love this one: Capital M looks like it was from a completely different font.

Then there’s this tiny request: Please add Monospace fonts for terminal use

This guy clearly showed up at the wrong party: 'y' and 'W' character appears too bold at 7 pt with Cleartype

But this one takes the cake. If you think Esperanto complaints are bad, try While we don't have a huge Klingon community in Ubuntu (at least not according to the Klingon Translations Team[1]), it would still be nice to at least be able to display the Wikipedia Klingon Language page[2] properly…

And people wonder why font designers don’t create more Open-Source fonts.

Khaled Hosny's picture

As an "open source" (I prefer to call it "free") font developer, I fail to see what is wrong with the issues you are citing, all are pretty valid concerns to me and should be put into consideration.

Jens Kutilek's picture

There is nothing wrong with most of the bug reports. I find most of them really interesting (true, there are some funny ones which show a lack of type design knowledge like the one James quoted about adding a Monospace family). It just seems to me that Linux has been flying under the radar of commercial font production until now*, but now a whole new can of worms has been opened.

On Windows the rendering options are quite simple: No smoothing, Greyscale smoothing, ClearType (+recently DirectWrite). That's 4 options.

On Ubuntu (Gnome? I kind of lost track there ...) there is No smoothing, Greyscale, Subpixel; with hinting ignored (Mac-like), slight, medium, and full. The interpreting of certain instructions may be switched on or off in FreeType (patent issues). That gives you over 20 combinations that may affect the final result.

* Yes, I know there are the Droid and Liberation fonts, but I haven't seen a bug tracker for them.

Khaled Hosny's picture

I don't see what is funny about asking for a monospaced font, Ubuntu font is intended to be an interface font, and a monospaced variant is a must. It might not be worded in the most type savvy way, but canonical is explicitly asking for community input which would certainly include requests like this (and they seem to acknowledge the need for a monospaced font). I agree about the diversity of the possible configurations and it might be a challenge to support all of those. But Ubuntu as an OS vendor (or distribution maker) can opt to only support whatever their default configuration is (plus another one or two most common configurations).

vinceconnare's picture

Most of the bugs seen are from unhinted pre-release or system/rasterizer or apps issues. Largest being the slight or medium settings which don't use the hints properly.

This is what it looks like on an Android mobile.

http://www.omgubuntu.co.uk/2010/10/new-ubuntu-font-meets-android-looks-s...

Other character set issues weren't in the design brief or are in process of development.

Richard Fink's picture

So I guess a small investment in time and money for Rosetta Stone/Klingon can really pay off.

Si_Daniels's picture

>On Windows the rendering options are quite simple: No smoothing, Greyscale smoothing, ClearType (+recently DirectWrite). That's 4 options.

>On Ubuntu ... That gives you over 20 combinations that may affect the final result.

Windows does come with a ClearType tuner that does let you adjust the ClearType to best suit your personal set of eyeballs. So there are more than 3 settings if you count the tuning subtleties.

Contrast with the zero options on latest Mac OS, down from 3 in previous version?

clauses's picture

I played around with the Ubuntu 10.10 RC today, and I must say that its rendering is – by far – the best between Windows, OS X and OSSs. OS X is far too fat/dark, Windows requires hinting and ClearType is (still) jagged.

I can't believe why Microsoft is still hanging on to the hinting approach. Apple on the other hand should take a clue from Ubuntu. Far superior fidelity to the outlines. And don't get me started on webbrowsers. Rendering quality must be at the bottom of their list.

John Hudson's picture

Clause: I can't believe why Microsoft is still hanging on to the hinting approach.

I can, because I get to see the clear benefits of hinting on a daily basis making fonts for a whole range of writing systems legible at 11 and 12 ppem. Since these are fonts primarily targeted at limited size ranges on screen, I get to design to the bitmaps as much as feasible (given that the fonts need to be reasonably future-proof against higher resolutions and zoomable environments), which in theory is the ideal situation for hintless rendering. But applying hinting to these size-targeted outlines has immediate, obvious visual benefits: stem density (blackness) is increased, stem edges are sharpened, small counters can be held open, letter proportions and sidebearings can be adjusted for proximate ppem sizes.

I'll go further: not only should people still be hinting, they should be using VTT to do it. FontLab's TT hinting is nice to have for mocking stuff up and for quickly checking GDI CT rendering during outline design, but the range of control and options available in VTT makes it pretty indispensable for high quality developing screen fonts.

clauses's picture

Okay, fair enough, but have you compared it to the current Ubuntu rendering? I'm curious to know how you rate it.

jasonc's picture

I'll go further: not only should people still be hinting, they should be using VTT to do it.

Here , here!

:)

John Hudson's picture

Claus, can you be more specific in terms of what I should be looking at on Ubuntu? What I've seen of text rendering on Ubuntu to date—reading over my wife's shoulder—has been inconsistent but tending towards the aenemic. I'm quite willing to believe that there is some good rendering happening somewhere on Ubuntu, but I'm not sure where to look for it. I have noticed wide inconsistencies between applications, e.g. between different browsers, so it would be helpful if you can precisely define the variables that result in what you think I should look at (Ubuntu version; add-ons or critical downloads; font flavour; application version; etc.).

malcolm's picture

John you’re absolutely right. Hinting for Ubuntu is like trying to hit a moving target.

The FreeType rasterizer on Ubuntu is configurable by any application, so you find that OpenOffice rasterizing is different from Mozilla Firefox rasterizing which is different again from the ‘standard’ system rastersing.

The Ubuntu system rasterizing is also extensively configurable by the user – from ‘full’ hinting down to ‘no’ hinting with switches turning off or on B/W, Greyscale and LCD. Some of these various levels have effects like ignoring ‘hdmx’ values and ‘cvt’ values.

Another problem here is that the fonts (under an OpenSource license) will inevitably end up on other OS’s so the hinting has to take account of this therefore the fonts cannot be hinted just for FreeType on Ubuntu.

clauses's picture

John, I've made some screenshots from Typekit's specimens http://typekit.com/fonts/adobe-text-pro That way others are able to pull up the same specimens on their computers for comparisons. Adobe fonts are famously auto-hinted, but if you want to point me to another font that will perform better in Windows let me know.

Waterfalls
Ubuntu 10.10 RC / Firefox
http://dl.dropbox.com/u/10940424/screenshots/Ubuntu%20Firefox.png

OS X 10.6.4 / Safari 5.0.2
http://dl.dropbox.com/u/10940424/screenshots/Safari%20OS%20X.png

Windows 7 / Internet Explorer 9 beta
http://dl.dropbox.com/u/10940424/screenshots/Windows%207%20IE9.png

Best renderings:
36 Ubuntu wins. OS X is too dark/fat. Windows is too thin, some strokes disappear (f, a, u, h, m, p). 'T' is flat on the top, compare with Ubuntu – I say it's better fidelity to the outline.
30 Id.
24 Id.
21 Id.
18 This is where the Windows (auto-)hinting falls apart. All the Windows renderings below this size are awful. Ubuntu still clearly in the lead, OS X is too dark.
16 Id.
14 Id.
13 From this size and down Ubuntu does become anemic compared to OS X, however it retains higher fidelity to the design, and it's not bad per se.
12 Id.
11 Id.
10 Id.
9 Id.

Sample text
Ubuntu 10.10 RC / Firefox
http://dl.dropbox.com/u/10940424/screenshots/Ubuntu%20Firefox%202.png

Smoothest rendering of the three. Keeps a contrast between vertical and horizontal strokes that OS X does not.
Major colour-fringing in negative text, but still the best between to three (but only because the three are generally horrible in negative).

OS X 10.6.4 / Safari 5.0.2
http://dl.dropbox.com/u/10940424/screenshots/Safari%20OS%20X%202.png

Blocky, splotchy (eg. vertical stroke in 'a') rendering with a noticeable lack of contrast between horizontal and vertical strokes.

Windows 7 / Internet Explorer 9 beta
http://dl.dropbox.com/u/10940424/screenshots/Windows%207%20IE9%202.png

What can I say? The Windows rendering is extremely jagged. Fidelity to the design is very low. People who prefer this should probably get prescription glasses or an LCD monitor, or both.

Comments?

Richard Fink's picture

@JH
"I'll go further: not only should people still be hinting, they should be using VTT to do it. FontLab's TT hinting is nice to have for mocking stuff up and for quickly checking GDI CT rendering during outline design, but the range of control and options available in VTT makes it pretty indispensable for high quality developing screen fonts."

Yes, people should still be hinting. For many reasons. After some testing with IE9 rendering, that's even more clear.

I disagree on the VTT part. Learning curve too high. Can't produce enough to meet the immediate demand. There are very acceptable, quicker alternatives.

Let's not get into a back-and-forth about what constitutes "high quality".
Today, my definition is "lack of distracting defects".
All of this IMHO.

@clauses
Do you know what format the Adobe font from Typekit is being delivered in?
Because when you say, "famously auto-hinted", that's PS hinting, not TrueType hinting.

rich

clauses's picture

I am under the impression that both TT and CFF OTF fonts of original Adobe designs are auto-hinted.

dberlow's picture

RF> Let's not get into a back-and-forth about what constitutes "high quality".

Okay. What constitutes auto-hinting then, IYHO?

And when one has to distinctively "hint" hundreds of fonts, the definition of quality perhaps goes beyond a lack of distracting defects, and onto containment of attractive effects?

Clauses> OS X 10.6.4 / Safari 5.0.2

(your first specimens of waterfalls), I don’t know where you got a Times on the Mac to dance with Century serifs at 36 pt. Is there some special song you play while rendering it?

malcolm> John you’re absolutely right. Hinting for Ubuntu is like trying to hit a moving target.

So, then... what's left to compare hinting for Ubuntu, Windows and the Mac OSs to again, I forget?

clauses's picture

David I don't understand your question. Care to rephrase that in prose?

John Hudson's picture

David: what's left to compare hinting for Ubuntu, Windows and the Mac OSs to again, I forget?

A moving target that's shooting back at you.

Richard Fink's picture

@db
"And when one has to distinctively "hint" hundreds of fonts, the definition of quality perhaps goes beyond a lack of distracting defects, and onto containment of attractive effects?"

Look, I know what you're saying. But what's the best that can be done for now? In bulk. Quickly.
As JH said, it's like trying to hit a moving target that's shooting back at you.
(Keep writing fellas, and I'll keep cutting and pasting.)

As of yesterday, the Typekit blog has started to hone in on sites using Typekit's library.

Sites We Like
One of them is Bigcartel.com

Using faces named LTC Bodoni and Ambroise.
Now, Ambroise, at least at the larger sizes used, looks OK. (There are a few defects - the capital H is missing it's crossbar, for example. At least on my machines.)

But LTC Bodoni is a disaster. At first, I thought some things - like the t's that overshoot the baseline were a design feature. A little bit of funk.
But no, it's just a crap hinting job all around. And it makes you wonder how something like that passes.
But it does. It do. And it's there for all to see. And pointed to with pride!

INFG wins again. And if crap sells, you have to ask yourself "why mess with a winning formula?"

Or maybe it's an opportunity for brand extension. LTC Bodoni Crap and LTC Bodoni Fixed.
And let the customer decide.

rich

John Hudson's picture

Rich: ...what's the best that can be done for now? In bulk. Quickly.

Not good enough.
_____

I disagree on the VTT part. Learning curve too high. Can't produce enough to meet the immediate demand. There are very acceptable, quicker alternatives.

Hinting is a design process, measured in the legibility of its results at specific sizes, so the notion of acceptability, like all notions of quality in type design, is directly related to individual typefaces for particular writing systems at specific sizes. There is no generalised basis for declaring one tool, process or strategy ‘very acceptable’, except insofar as the tool enables one to do everything that might need to be done to achieve the desired results.

If you go back and read the post in which I made the comment about VTT, you'll understand that I'm talking specifically about situations in which autohinting or even manual hinting in FontLab is not sufficient, because that tool doesn't enable one to do everything that needs to be done. Claus was wondering why Microsoft still rely on hinting in their rasterisation model, instead of ignoring hints in the way that Apple do or some applications using Freetype do. I replied with reference to the visible benefits of hinting generally, and more specifically in context of making dedicated screen fonts for small ppem sizes for writing systems that have much more stroke complexity than Latin or other European scripts. I might suggest that the reason Microsoft still include hinting in their rasterisation approach is that they have a much better understanding of the needs of internationalisation than any other OS developer. There are plenty of glyphs in plenty of writing systems that, without hinting control, resolve into grey blobs at typical text sizes on screen.

dberlow's picture

Rich> And if crap sells, you have to ask yourself "why mess with a winning formula?"

I’d ask if only crap sells. But the winning formula you refer to, is that there is no one winning formula, because there are many more than just one baby, baby.

Rich> I disagree on the VTT part. Learning curve too high.

I think that’s s’posed to be “too steep,” as only an overshooting curve can be too high. And, perhaps VTT is documented for another age and some things have happened to change the curve steepness that are not reflected either in its operation or its documentation? I’ll ask.

John>

Your suggestion, that Microsoft still include hinting because they have a much better understanding of the needs of internationalization than Apple e.g., I think is partly not on target. Microsoft still include hinting because they have less control over fonts, font software and display devices as a percentage of their OS base vs. Apple who is in control of a huge percentage, (down to the very fonts allowed in most of their new devices).

>There are plenty of glyphs in plenty of writing systems that, without hinting control, resolve into grey blobs at typical text sizes on screen.

And so they zoom, (the old fashioned way) on most Macs, which I guess one can’t just assume on all Windows devices? I think the issue is not, with all due respect, understanding of the needs of internationalization so much as solving the needs of open-system font adoption on an international scale, without control over the whole process from what font is loaded, to pixels on the screen.

And this, as you know I believe, is a situation without a process adequate to manage the unfolding complexities of a web-gone-property-wild.

John Hudson's picture

David: I think the issue is not, with all due respect, understanding of the needs of internationalization so much as solving the needs of open-system font adoption on an international scale, without control over the whole process from what font is loaded, to pixels on the screen.

Yes, that's fair and more nuanced analysis. Thanks.

Re. zooming: I'm having to think a lot these days about UI fonts, localisation, and screen real estate, so I don't see zooming as an adequate solution to screen legibility issues. I really wish it were, because I'm going a bit mad trying to figure out ways to make some glyphs legible at 11ppem.

Looking forward, the answer has to be adaptive layout that is sensitive to locale settings and language of everything from the UI to web/document content. But at the moment, UIs tend to be immutable and, alas, often designed around English text strings and font metrics.

brianskywalker's picture

> Klingon

So has anyone seriously explored adaptations of the Klingon alphabet that will work well on the same line as the Latin alphabet? Perhaps there is some way find the elemental form of the characters and create new typefaces based upon that. One must wonder if there are display and text varients of the language, or how their letterforms developed. Was there a traditional calligraphy? It looks like the letterforms may be easily written with a brush.

While the amount of fluent Klingon speakers in the world is small, the ones that do use this language will obviously appreciate the development. And, if you don't think Klingon is a serious language, consider that you can get Hamlet in Klingon. And, in the words of a great Klingon:

"You can't appreciate Shakespeare until you've read him in the original Klingon."
General Chang

John Hudson's picture

While the amount of fluent Klingon speakers in the world is small, the ones that do use this language will obviously appreciate the development. And, if you don't think Klingon is a serious language, consider that you can get Hamlet in Klingon.

Humans who study, read and write the Klingon language do so almost exclusively in a Latin script alphabet, not in the Klingon script. It was on this basis that the Klingon script was rejected for encoding in Unicode: there was no expressed desire or need from the user community.

brianskywalker's picture

> Humans who study, read and write the Klingon language do so almost exclusively in a Latin script alphabet, not in the Klingon script.

Oh snap.

And I did find a nice interpretation of Klingon....
http://www.typotheque.com/blog/typotheque_fonts_now_support_klingon

Unfortunately, it was posted on April 1st

dberlow's picture

John> I don't see zooming as an adequate solution to screen legibility issues.

computerized zooming and human zooming are the most popular means of changing the font size now for over 2000 years, good luck!

John> ...the answer has to be adaptive layout that is sensitive to locale settings and language of everything

...but not adaptive, or adaptive enough fonts.

Richard Fink's picture

@jh

>>Rich: ...what's the best that can be done for now? In bulk. Quickly.

>Not good enough.

It's gotta be and besides, it will be. But that doesn't mean there shouldn't levels.
There is the mass-produced and the hand-tailored. Remember, too, that a large part of the market using Mac-iPhone-iPad/Quartz aren't going to see the results anyway no matter how hard anybody slaves away.

I've been spending a lot of time in FontLab over the past month and one of the things I focused on was learning to manually hint. The documentation - which I've read over and over - isn't bad. I've gotten info from other sources too - whatever I could find. But I have to assume something's missing because my brain remains unpenetrated. I'm still largely mystified by the results I get.
But that's OK, because it at least answered the question for me: "How accessible is this stuff?"
I've put it aside and will revisit - much else to do with bigger payoffs.

At this point, I would imagine MSFT sees it as against their own interest to start offering more intuitive font-creation tools. They can get what they need from the existing pool of cogniscenti familiar with their ways, and that's that. Why get more people involved - to create better quality fonts for Google and Sun to wield as hatchets as they chip away at MS Office? I don't think so.

>I don't see zooming as an adequate solution to screen legibility issues.

Zoom, zoom, zoom. Interestingly, it has gone completely unremarked anywhere as far as I know that IE9 has done away with IE's Text Size menu. Gone *and* instantly forgotten. Resizing text independently of the surrounding content is now totally in the realm of on-page JavaScript widgets. [A+] [A-]
I'd rather have seen a Text Only resize/reflow remain in the browser chrome but these days on mobile platforms there's nowhere in the chrome to put the feature. And there's no spec that allows you to detect - cross browser/cross platform - when the user zooms so as to present an alternative. (I've done a lot of work on this - to completely no avail.)
So here we are, paddling on.

Syndicate content Syndicate content