WTF site



  • http://www.aiseikai.or.jp/ - In case you haven't seen it yet.

    I wouldn't like to land in that hospital. Seems like they may be familiar with miko miko nurse. (NSFW... if you have sound turned on).



  • I'm going to have nightmares about this...



  • I think this is made so you have a seizure (sp?) and have to go to this hospital.



  • X had a seizure when that first link loaded. It crashed... I dare not try the second link.



  • The Real WTF™ is those bastards copied my site design...


     

     



  • Just realized, that in this picture: 

    the rotating red character looks like bloody axe... maybe that's just in my head...



  • Thank God the site is copyrighted, because you wouldn't want people stealing that stuff for their own sites.



  • I love the scrolling message across the top, which proudly explains that the site was made with FrontPage, and goes on to detail (very slowly) exactly what won't work in browsers other than IE6 at 800x600 with font size set to "medium".

    I hope the person who perpetrated this monstrosity has done the honourable thing to atone for his crime.



  • Dear god, let the designer of this site achieve one of the Darwin awards.



    And I thought that I am really bad designer.....
     

     



  • Oh my god. That is amazing. I love it!



  • Is it just me, or is the real WTF Japanese and Chinese fonts?

    They always contain a copy of the Latin alphabet for International text, but in Windows at least, it's always hideous. In that example, the letters don't even all sit on the baseline, and they're not hinted properly leading to strange and irregular letter shapes. Nor are they correctly kerned. Even in printed product manuals the Latin letters look ugly.

    I don't know why it's too much of a problem to correctly and aesthetically draw and hint 52 Latin letters in a typeface that has thousands of Chinese or Japanese characters in. It's only 52 more and they're well-documented shapes. (Well, you'd also want to do the Arabic-Indic numbers and Latin punctuation too.)

    Unicode and international text is a weird mess. Computers only understand typefaces, and plenty of UTF-aware apps still drawn a square or nothing for characters not found in the selected font. Web browsers are some of the few if not the only apps capable of fetching characters from other fonts as necessary either for language reasons or for symbols. I don't see why Chinese and Japanese fonts need to bother with defining the Latin alphabet -- just have the software grab them from an appropriate Latin font (serif or sans-serif, probably Times or Arial).



  • @Daniel Beardsmore said:

    Is it just me, or is the real WTF Japanese and Chinese fonts?

    They always contain a copy of the Latin alphabet for International text, but in Windows at least, it's always hideous. In that example, the letters don't even all sit on the baseline, and they're not hinted properly leading to strange and irregular letter shapes. Nor are they correctly kerned. Even in printed product manuals the Latin letters look ugly.

    The Windows font selection and rendering algorithms are known to suck utterly for non-English text, and particularly for text that has to mix fonts. The fonts I use on X are somewhat better, although I still don't have a perfect match (Eastern fonts are oddly shaped, and their presence in any text has significant effects on the metrics of all the rest of the text).

    TeX gets font rendering right. So does QuarkXPress when targeting a real printing press. Everything else gets it wrong.

    I don't see why Chinese and Japanese fonts need to bother with defining the Latin alphabet -- just have the software grab them from an appropriate Latin font (serif or sans-serif, probably Times or Arial).

    Precisely because the Windows multi-font algorithms suck.



  • Actually, in most Chinese and Japanese fonts the latin letters are <font face="MS Gothic">monospaced</font>, and for some commonly used sizes, the fonts also contain bitmaps - which makes the text look jaggy. <font face="MS Gothic">And of course, don't forget about the full-width letters.</font>



  • @ender said:

    Actually, in most Chinese and Japanese fonts the latin letters are <font face="MS Gothic">monospaced</font>, and for some commonly used sizes, the fonts also contain bitmaps - which makes the text look jaggy.

    Bitmap fonts only make text look jaggy when the font rendering algorithm is criminally broken in the most fucktarded way imaginable.

    Windows, I'm looking at you. 



  • @ender said:

    Actually, in most Chinese and Japanese fonts the latin letters are <font face="MS Gothic">monospaced</font> ...

    This is logical given that Chinese and Japanese text is monospace, and this stops Latin text from skewing all their own text although that may not be the precise reason for it. It also has the effect of looking quite hideous, giving you a choice of v e r y w i d e l y s p a c e d l e t t e r s or all the monospace character columns out of line with each other due to the insertion of Latin text. I imagine the latter is far preferable but I have no idea.

    @asuffield said:

    Bitmap fonts only make text look jaggy when the font rendering algorithm is criminally broken in the most fucktarded way imaginable.

    Windows, I'm looking at you.

    I'm curious what you feel Windows font rendering does to break bitmap fonts. I always just felt that the Latin letters in Chinese and Japanese fonts were just drawn badly. I'm sure you remember the hideous plotter fonts in Win16 that were Microsoft's best efforts at scalable type before they adopted TrueType -- that's what I'm reminded of :) Although the plotter font rendering was its own level of hideous and unreadable.

    Another bitmap font that sucks rocks is the WST_* series, for World System Teletext. The letters don't remotely resemble British Teletext on which they're based. It looks like the Teletext symbolic and graphic sections (the Game of Life creatures and Tetris blocks) was just slapped onto some random hideous font, mis-proportioned, mis-drawn and one that Character Map can't even render correctly when you try to preview it. Yummy.



  • Incidentally, the Wikipedia page on Teletext has a screenshot of Teletext that gets the character set correct, but fails to draw correct double-height text. Double-height text is more than just doubled pixel rows, it's dedicated larger type, although the hardware anti-aliasing in the SAA 5050 chip is a bit iffy at that size.



  • @asuffield said:

    @ender said:

    Actually, in most Chinese and Japanese fonts the latin letters are <font face="MS Gothic">monospaced</font>, and for some commonly used sizes, the fonts also contain bitmaps - which makes the text look jaggy.

    Bitmap fonts only make text look jaggy when the font rendering algorithm is criminally broken in the most fucktarded way imaginable.

    Windows, I'm looking at you. 

    Or if the person claiming that this is the case is of the opinion that non-antialiased text is inherently "jaggy" (i assume your 'broken' claim relates to scaled bitmaps; but there are many people who think that even unscaled bitmaps are 'jaggy')



  • @Daniel Beardsmore said:

    Incidentally, the Wikipedia page on Teletext has a screenshot of Teletext that gets the character set correct, but fails to draw correct double-height text. Double-height text is more than just doubled pixel rows, it's dedicated larger type, although the hardware anti-aliasing in the SAA 5050 chip is a bit iffy at that size.
    That depends a lot on the decoder - just about every TV I've seen renders the teletext slightly differently. I don't think I've seen any TV antialias the letters, although the older ones certainly appeared to do so, but that was because the picture was smoother in general (they had the exact same effect on PS2 games). The double (and quad, when using zoom-in) height letters on most seem to be just doubled pixel rows on most TVs, although our new Samsung does seem to actually use a different type.@Daniel Beardsmore said:
    It also has the effect of looking quite hideous, giving you a choice of v e r y w i d e l y s p a c e d l e t t e r s or all the monospace character columns out of line with each other due to the insertion of Latin text.
    The very widely spaced letters are actually full-width letters, occupying Unicode positions 0xFF01 to 0xFF5E. With the normal ("half-width") latin letters, you get 2 of them per one Chinese character.@Random832 said:
    Or if the person claiming that this is the case is of the opinion that non-antialiased text is inherently "jaggy"
    That's exactly what I had in mind.



  • Aaarrgh!!! I tried to edit my first post above, and it wouldn't let me (edit timeout). Sooo ... I edited my second post instead, as I'd just made that. I hit Submit and then went off to do something else (my Web access has been dreadfully slow lately). Came back after dinner/TV etc to find that although Community Server let me start editing the post, it waited until I'd composed an edit to ban me from editing it. Came back to a nice ban message.

    My correction was, that Character Map does correctly render the WST_* fonts, but they're drawn even more incorrectly than I realised!

    @ender said:

    That depends a lot on the decoder - just about every TV I've seen renders the teletext slightly differently.

    I've never seen a TV deviate from the standard typeface that goes back to when Teletex was invented.

    @ender said:

    I don't think I've seen any TV antialias the letters, although the older ones certainly appeared to do so, but that was because the picture was smoother in general (they had the exact same effect on PS2 games).

    It's part of the 1982 Mullard SAA5050 generator chip specification, the Teletext ASCII generator used on the BBC Micro. What generator chip and process other decoders use, I don't know.

    It's a dynamic process, leading to some oddly shaped characters such as the top of a '4', where the diagonal smoothing stops one pixel short of the top. From the specification, "The alphanumeric characters are character rounded, i.e. a half dot is inserted before or after a whole dot in the presence of a diagonal in the character matrix."

    But given the general fuzziness of the signal, even on an old Microvitec monitor, it's really hard to figure out how much of the blur is the smoothing pixels and how much is just an out of focus image. This makes double-height characters extremely hard to figure out, although it's probably in the spec. It's 16 separate TIFF images, which need putting together into a single PDF. I imagine my Mac could do that if I knew how.

    @ender said:

    @Random832 said:
    Or if the person claiming that this is the case is of the opinion that non-antialiased text is inherently "jaggy"
    That's exactly what I had in mind.

    Not at all. Windows makes extensive use of non-antialiased text and it looks good. Clear, readable, attractive. But the Latin letters in Japanese and Chinese fonts are just horrible, really badly drawn, or simply badly hinted or lacking hinting at all. It's the hinting that allows scalable type to be rendered nicely without anti-aliasing and you may find that this is where the problem lies.



  • @Random832 said:

    @asuffield said:

    @ender said:

    Actually, in most Chinese and Japanese fonts the latin letters are <font face="MS Gothic">monospaced</font>, and for some commonly used sizes, the fonts also contain bitmaps - which makes the text look jaggy.

    Bitmap fonts only make text look jaggy when the font rendering algorithm is criminally broken in the most fucktarded way imaginable.

    Windows, I'm looking at you. 

    Or if the person claiming that this is the case is of the opinion that non-antialiased text is inherently "jaggy"

    Well, there's no accounting for blind retards who don't know what the English alphabet looks like. Every time I see somebody complaining that an "E" looks jagged I feel a sudden urge to stab them in the eyes - it's four straight lines, what the hell do you expect? It's supposed to have corners.

     

    (i assume your 'broken' claim relates to scaled bitmaps; but there are many people who think that even unscaled bitmaps are 'jaggy')

    Yes, fractional bitmap scaling is just idiotic. Don't do it. Ever. Bitmap fonts can be scaled by exact integer multiples of pixels and nothing else. If a font renderer is instructed to render a bitmap font at 13 points and it only has a 12 point version, then some kind of back-arsewards fractional scaling is only going to make a horrendous mess, because there is no way that it is going to fit onto the raster grid of your display. There is no such algorithm that can take n lines of black and white and turn them into n+1 lines that still looks the same - it simply isn't possible, for all values of n > 2. The best thing to do is draw the 12 point font on the screen and leave extra white space between them (still ugly, but nowhere near as bad as attempts to scale bitmap fonts to fractional sizes), and/or tell the user to stop asking for fonts that don't exist. If you want arbitrary scaling, use a vector font. If you want crystal-clear rendering all the time, use metafont, or use a good quality bitmap font and don't scale it.

    Unscaled bitmaps can still be jaggy if you use crap fonts that are actually jaggy. Most of the bitmap fonts shipped with Windows are (historical) crap. This doesn't help.

    By their very nature, it is impossible for any font to be better than an unscaled bitmap font on contemporary software, because every other kind is converted into a bitmap font by the renderer, when it draws them to the screen. You cannot have higher quality than the format used by your output device, and your display is a bitmap.

    Hrrrngh. 



  • @asuffield said:

    @Random832 said:
    @asuffield said:

    @ender said:

    Actually, in most Chinese and Japanese fonts the latin letters are <font face="MS Gothic">monospaced</font>, and for some commonly used sizes, the fonts also contain bitmaps - which makes the text look jaggy.

    Bitmap fonts only make text look jaggy when the font rendering algorithm is criminally broken in the most fucktarded way imaginable.

    Windows, I'm looking at you. 

    Or if the person claiming that this is the case is of the opinion that non-antialiased text is inherently "jaggy"

    Well, there's no accounting for blind retards who don't know what the English alphabet looks like. Every time I see somebody complaining that an "E" looks jagged I feel a sudden urge to stab them in the eyes - it's four straight lines, what the hell do you expect? It's supposed to have corners.

     

    (i assume your 'broken' claim relates to scaled bitmaps; but there are many people who think that even unscaled bitmaps are 'jaggy')

    Yes, fractional bitmap scaling is just idiotic. Don't do it. Ever. Bitmap fonts can be scaled by exact integer multiples of pixels and nothing else.

    In this case, I'm now entirely at a loss for what you think Windows is doing wrong, since it refuses to scale by anything other than an exact multiple.


    By their very nature, it is impossible for any font to be better than an unscaled bitmap font on contemporary software, because every other kind is converted into a bitmap font by the renderer, when it draws them to the screen.

     

    You cannot have higher quality than the format used by your output device, and your display is a bitmap.

    Maybe you're still on an old MDA, but the rest of us have moved on.

    And, a non-antialiased "/" character is still going to be made of several rectangles joined at the corners. (or, in some cases, joined at the top and bottom with a horizontal offset) There's nothing any bitmap font can do to change that. Nobody ever complains about how "E" looks; that was a straw man argument.



  • @Random832 said:

    You cannot have higher quality than the format used by your output device, and your display is a bitmap.

    Maybe you're still on an old MDA, but the rest of us have moved on.

    And, a non-antialiased "/" character is still going to be made of several rectangles joined at the corners. (or, in some cases, joined at the top and bottom with a horizontal offset) There's nothing any bitmap font can do to change that. Nobody ever complains about how "E" looks; that was a straw man argument.

    Kindly describe this magical way for rendering a "/" that cannot be represented by a bitmap. 



  • Pen plotter.



  • @Daniel Beardsmore said:

    Unicode and international text is a weird mess. Computers only understand typefaces, and plenty of UTF-aware apps still drawn a square or nothing for characters not found in the selected font. Web browsers are some of the few if not the only apps capable of fetching characters from other fonts as necessary either for language reasons or for symbols. I don't see why Chinese and Japanese fonts need to bother with defining the Latin alphabet -- just have the software grab them from an appropriate Latin font (serif or sans-serif, probably Times or Arial).

    I've recently been busy with this, and implemented just that, if it can't find the character in the font then it uses a other font to find it. (if that doesn't work, then i'll just show the original fonts not-found glyph)

    Currently i'm using the Dejavu font because it has a massive shit-load of characters. 

    But it's a bit of a bitch to do, currently i loop over all characters in the text and cross reference that to a dump from the font. It's fast enough, but far from optimal. Especially since i would recon this is typically something a library should be doing, however the library i'm using only supports defining a replacement character for the character that isn't found.

    But currently it's working and now i got a whole new level of problems. like mixed left to right and right to left characters. *sigh* The library fucks it up, and mirrors the right to left characters.

    Although i have to admit, all this is done in PHP, so what should i even expect. *sigh* although i found that native unicode support will be in PHP6. Little help now though.
     



  • @asuffield said:

    Kindly describe this magical way for rendering a "/" that cannot be represented by a bitmap. 

    Render it as an image with pixels of more than one bit (and thus, is not a bitmap). (yeah, yeah, they're commonly called bitmaps, but "bitmap fonts" universally refers to fonts consisting of real bitmaps)



  • @Random832 said:

    @asuffield said:

    Kindly describe this magical way for rendering a "/" that cannot be represented by a bitmap. 

    Render it as an image with pixels of more than one bit (and thus, is not a bitmap). (yeah, yeah, they're commonly called bitmaps, but "bitmap fonts" universally refers to fonts consisting of real bitmaps)

    I dispute the universality of this definition.

    Fonts are almost always greyscale, so the standard technique for handling these is to oversample the stored image: if you want to store a 12x12 box with a depth of 4 bits, then you store it in 24x24 pixels, with each 2x2 region representing one pixel; when rendered into a 12x12 space you get the expected image. Just because the file format doesn't include explicit depth doesn't mean it isn't possible. This technique is most notably used by metafont and TeX (so it's been around for about 20 years).
     


Log in to reply