| | <<<Back 1 day (to 2011/09/22) | 2011/09/23 |
| mvrhel2 | nice! | 00:01.30 |
| | henrys: I am sending out an email about my lunch with MQ | 00:01.46 |
| kens | Well, after some effort, I've convinced myself that the only way to 'properly' handle Multiple Master fonts is going to be to write a complete parser. The Adobe fonts manipualte teh stack, and call subroutines, so the only way to ensure teh numbers being used are correct is to actually run the CharString. | 08:32.12 |
| chrisl | And how much work would it be to allow pdfwrite to gather vector data for Type 3 fonts? | 08:33.27 |
| kens | Type *1* fonts. | 08:33.39 |
| | It already captures type 3 fonts by executing the BuildChar | 08:33.53 |
| | But type 1 fonts don't have a BuioldChar | 08:34.02 |
| chrisl | I meant gather vector data *from* any font type, to go *into* a Type 3 | 08:34.09 |
| kens | That would be an immense amount of effort. | 08:34.23 |
| | In addition to a type 1 pareser I'd need a CFF and Type 42 pareser. | 08:34.35 |
| | Or at least a way of extracting an outline from said fonts. | 08:34.48 |
| chrisl | I'd hoped you could intercept the path before the glyp cache | 08:35.18 |
| | ahem - glyph cache renders it..... | 08:35.31 |
| kens | I don't think that's possible, we don't get a path do we ? | 08:35.36 |
| | FreeType just renders the bitmap for us. | 08:35.46 |
| | I have something which 'mostly' works by just using the 2 bottom numbers on the stack. | 08:36.18 |
| | These are the initial values that get modified by the $Blend | 08:36.32 |
| | SO by always using those two we get teh same basic font. | 08:36.45 |
| chrisl | Well, we could tell freetype to give us the path instead of rendering, but it sounds like it would end up as more work that your current approach. | 08:37.10 |
| kens | It would probably be easier, but I don't think we currently have a font method that says 'give me the path for this glyph' | 08:37.47 |
| | And even if we did, I'd still have to cast it back into Type 1 font operations. | 08:38.04 |
| | Oh, and hinting would go. | 08:38.28 |
| | Not that that's a problem in this case. | 08:38.40 |
| | I'm inclined to go with my current code, as long as it doesn't cause regressions. Its pretty unusual to use a MM font, and be producing PostScript. | 08:39.37 |
| chrisl | I hadn't thought you'd recreate a Type 1, as I said, just a vector Type 3 - it would mean work outside pdfwrite, though. | 08:40.01 |
| kens | I suppose that would be possible, no more ugly than what we have now. | 08:40.31 |
| | Seems like a lot of work though. | 08:40.39 |
| chrisl | Yes, probably. | 08:40.52 |
| kens | I've started a regression test with this code, if it works I'm going to commit it with a bunch of comments saying what else would have to be done. | 08:41.12 |
| | Im *really* fed up of this now. | 08:41.33 |
| | BTW this explains why the CFF conversion works, it really does fully interpret the glyph at the time of conversion. | 08:41.59 |
| chrisl | Yes, well I'm pretty fed up with CIDFont substitution........ | 08:42.12 |
| kens | So if I was going to do this, then I would start agin by leveraging off that code, and having it produce type 1 instead of CFF | 08:42.34 |
| | (optionally) | 08:42.43 |
| chrisl | Makes sense - depending on how horrific that code might be! | 08:43.18 |
| kens | Its horrific, which is why I didn't want to do it in the first place, but now I understand the problem, its the only way to go. | 08:43.41 |
| | Making it flexcible enough to produce T1/CFF wouldn't improve its readability though ;-) | 08:44.01 |
| Robin_Watts_ | marcosw_: You alive? | 09:28.50 |
| | kens: Is git cluster working for you at the moment? | 09:33.35 |
| | oh, ignore me. | 09:34.37 |
| kens | Yes, oh,... | 09:34.48 |
| | chrisl my code shows progressions in 2 test suite files, so I'm going to commit ti with large warnigns about 'not a complete solution' | 09:35.42 |
| chrisl | Well, that's good. Does the pdfwrite output now match the rendered output? | 09:41.08 |
| kens | ps2write. | 09:41.21 |
| | No, but it does now produce legible output, whereas previously it was blank | 09:41.37 |
| chrisl | Sorry, ps2write | 09:41.38 |
| kens | Its 'better' but not 'right' | 09:41.53 |
| chrisl | Close enough! | 09:42.03 |
| kens | It will have to do for now, this is taking *way* too much time. | 09:42.18 |
| | Big comment going in at the head of the code. | 09:42.29 |
| kens | needs more caffeine.... | 10:11.21 |
| Robin_Watts_ | Bah. Looks like I need to code another strip_copy_rop planar 4bit case. | 11:46.25 |
| | I have a file that used to run 5 times as slow on plank as with pamcmyk4. I've just special cased a bit of code, and now it runs 3 times FASTER. | 14:13.14 |
| | And the output results are identical. | 14:13.20 |
| | That's got to be a mistake somewhere... | 14:13.26 |
| henrys | fixing the text color space in pcl is going to be more work than expected but should improve performance quite a bit. | 14:28.05 |
| | Friday - yeah! | 14:31.23 |
| Robin_Watts_ | Did the discussion yesterday end up thinking it should fixed in PCL rather than in the gfx lib ? | 14:35.52 |
| henrys | I really should do it in pcl - I've started, should have something early next week, it will change every character in the tests ;-( | 14:38.25 |
| kens | marcos will be delighted | 14:38.48 |
| henrys | I know I was talked out of removing libpaper but I can't remember why. | 14:53.56 |
| Robin_Watts_ | Item 5 last meeting agenda but one, get rid of libpaper | 14:55.18 |
| | My notes say: "Done" | 14:55.29 |
| chrisl | How would we do paper sizing on Unix systems? | 14:55.48 |
| Robin_Watts_ | The libpaper fix may be as simple as removing the free call. | 14:56.52 |
| chrisl | Then we leak memory, I think. | 14:57.11 |
| henrys | why is unix paper size different than windows paper sizing, again? | 14:57.13 |
| chrisl | Unix doesn't guarantee having the same localization settings as you get on Windows. | 14:57.58 |
| Robin_Watts_ | chrisl: Well, either systempapername is returning a pointer to an alloced buffer, and we should free it,... or it's not, and we shouldn't. | 14:58.58 |
| | The bug would seem to imply the latter. | 14:59.04 |
| chrisl | The argument was we should either use the recommended paper size selection (where available) for a given platform, or we should *always* do our own "something". The feeling was Windows users wouldn't like that. | 14:59.08 |
| henrys | ah right, but wasn't that like 6 lines of code that we could easily just absorb into gs. | 14:59.19 |
| chrisl | I don't think we have the information on Unix to use that 6 lines of code..... but that's from memory | 15:01.00 |
| | Also, I think we'd get hammered by the distribution maintainers if we took out libpaper. | 15:01.34 |
| | I guess we could make it an "opt-in" feature, but then it might rot..... | 15:02.18 |
| henrys | but our customers can't use it. | 15:02.18 |
| | so customers are missing the functionality. | 15:02.40 |
| Robin_Watts_ | My memory was that we were going to absorb the 6 lines of code from windows into gs, and offer libpaper as a compile option. | 15:03.44 |
| | but my memory is poor. | 15:04.00 |
| henrys | be back in an hour or so. | 15:16.15 |
| Robin_Watts_ | marcosw_: Is plank enabled for pcl on the cluster yet ? | 15:19.02 |
| marcosw_ | I don't recall, let me check. | 15:19.19 |
| Robin_Watts_ | I bet it's not. | 15:19.25 |
| marcosw_ | why do you say that? | 15:19.41 |
| | it's not. do you want me to enable it? | 15:20.02 |
| Robin_Watts_ | not yet. | 15:20.12 |
| | I'll get 1 more commit in, then can we run a pcl plank vs pamcyk4 difference check please? | 15:20.31 |
| marcosw_ | sure. | 15:20.39 |
| | do you want fast thresholding turned on? | 15:21.00 |
| Robin_Watts_ | No. | 15:22.08 |
| | mvrhel2: Did you send the email about your lunch with Mq ? | 15:25.26 |
| | marcosw_: OK, could you run that test now please? (differences between plank and pamcmyk for pcl, no fast thresholding. Might as well make it 300dpi banded only, I guess) | 15:40.00 |
| marcosw_ | will do. | 15:40.10 |
| Robin_Watts_ | Thanks. | 15:40.13 |
| | marcosw_: Actually, scratch that test. | 15:47.30 |
| | I've just spotted another bug. | 15:47.37 |
| marcosw_ | okay. | 15:47.39 |
| Robin_Watts_ | Sorry. | 15:47.42 |
| marcosw_ | I'm going to be offline for an hour or so, send me email when you want me to start the plank vs. pamcmyk4 test. | 15:53.17 |
| mvrhel2 | Robin_Watts: I did, I think I forgot to cc you on it sorry about that. resending | 15:58.45 |
| | hmm. I know I sent this thing oug | 16:00.26 |
| | out | 16:00.28 |
| | ha. forgot to hit the send button | 16:02.55 |
| | you were on the list Robin_Watts | 16:03.03 |
| | I was wondering why no one had replied to my email | 16:03.15 |
| chrisl | Phew, another cups related problem bites the dust :-) | 16:04.34 |
| kens | chrisl it seems to be calling cshow, which is why it doesn't go through the usual show routines :-( | 16:17.06 |
| | Now I have to find out where it is going.... | 16:17.14 |
| chrisl | kens: well, at least you know | 16:20.37 |
| kens | :-) | 16:24.19 |
| | Alwyas look on the bright side of life.... | 16:24.33 |
| chrisl | Yes, indeed! | 16:25.16 |
| Robin_Watts_ | mvrhel2: ixwebhosting was DDOS attacked yesterday, so I wondered if I'd missed it :) | 16:42.39 |
| mvrhel2 | I don't know how I managed not to hit send | 16:43.05 |
| Robin_Watts_ | Read it now. Interesting. | 16:44.04 |
| | Really annoying as I'd done no target specific optimisation yet, so I bet we could have come out the winner in terms of speed! | 16:44.41 |
| mvrhel2 | yes. I was a bit surprised about that | 16:45.58 |
| | and the price issue is also bizarre | 16:46.15 |
| | It all really makes no sense to me | 16:46.33 |
| Robin_Watts_ | Bonkers. I'm getting a strip_copy_rop with a texture width and height of 1. | 17:24.16 |
| henrys | mvrhel2:yes the monotype/pcl situation is very bad for us, I was waiting for it to start effecting us and it has. | 17:32.19 |
| mvrhel2 | ok. I was not aware that they had pcl until now | 17:32.42 |
| ray_laptop | I'm also curious about monotype's PCL. henrys, did you know they had one or know anything about it ? | 17:38.15 |
| henrys | I had heard it was in development, I don't see how we can compete if we can't get folks to take up our urw font solution. | 17:39.23 |
| | customers pay more for the ufst fonts than they do for our interpreter. | 17:40.16 |
| mvrhel2 | ok that is not going to work | 17:40.38 |
| Robin_Watts_ | Are monotypes fonts seen as being "better" then? | 17:41.11 |
| ray_laptop | if companies already have UFST licensed, then that is 'sunk cost' -- I can't believe that we can't compete on PDL's (particularly against a partial Adobe situation), and what about XPS -- does MI have that as well ? | 17:41.26 |
| | Robin_Watts_: they are smaller in ROM footprint, but generally the font quiality is inferior | 17:41.48 |
| mvrhel2 | apparently they have xps looking at their website | 17:41.53 |
| henrys | ray_laptop:that's exactly backward they have patented compression we don't have the quality is about the same. | 17:42.27 |
| | oh I read it backwards nvm. | 17:42.40 |
| Robin_Watts_ | ray_laptop: ufst fonts are smaller than (but inferior to) our urw offerings, yes, sorry I knew that. | 17:43.01 |
| henrys | they also have trademark names try selling nimbus foobar vs. courier | 17:43.13 |
| ray_laptop | Robin_Watts_: but since the fonts are _so_ much smaller, everybody uses them, so their isn't much concern over the qualtiy -- particularly at 600+ dpi | 17:43.39 |
| mvrhel2 | hmm. I wonder what is so magical about their compression scheme | 17:44.05 |
| Robin_Watts_ | Monotype == ufst based ? | 17:44.17 |
| mvrhel2 | http://ir.monotypeimaging.com/releasedetail.cfm?ReleaseID=489819 | 17:44.28 |
| | this is an old press release | 17:44.36 |
| henrys | if monotype plays their cards right which I suspect they will pcl will be bundled with the ufst and upgrades will have dependencies. | 17:44.45 |
| ray_laptop | MANY printer companies have been successful with "Fonts simulating Courier" kind of statements | 17:44.49 |
| henrys | yes I know that but still many companies won't switch - try having the conversation with Norbert. For those companies I am concerned about our pcl business. | 17:46.14 |
| ray_laptop | but I don't see the business argument for monotype -- they already 'own' the font market, at least for the low end where ROM size is an issue, and if they 'give away' their PDL's for the same cost as the fonts+UFST alone, they are not increasing revenue and profit will be less due to support and development of the PDL | 17:47.35 |
| Robin_Watts_ | henrys: So the major disadvantage of our fonts is the naming? | 17:47.49 |
| | Not the compression? | 17:48.04 |
| ray_laptop | Robin_Watts_: and the rom size | 17:48.06 |
| | our compressed URW set of 80 is still about 3Mb -- theirs is less than 1Mb | 17:48.38 |
| mvrhel2 | I wonder if they are using some visual based compression method | 17:49.09 |
| Robin_Watts_ | ray_laptop: Right, but if that's a major factor, we just do our own font format that's more compact. | 17:49.11 |
| mvrhel2 | yes | 17:49.18 |
| henrys | It is hard for me to believe that norbert has issue with rom footprint with the monstrosity they are selling but it could be. | 17:49.33 |
| ray_laptop | mvrhel2: their compression shares partial edges of glyphs | 17:49.34 |
| mvrhel2 | ray_laptop: so you mean they chop things up into common parts? | 17:50.11 |
| ray_laptop | mvrhel2: so the right edge of a D might be shared with part of an O | 17:50.12 |
| mvrhel2 | yes | 17:50.15 |
| | ok | 17:50.18 |
| | makes sense | 17:50.25 |
| Robin_Watts_ | We could do that. Do you know the patent number? | 17:50.36 |
| ray_laptop | mvrhel2: I think it was all 'hand tweaked' | 17:50.50 |
| mvrhel2 | sure | 17:50.59 |
| Robin_Watts_ | Sorry, to be clear, the compression is "Monotype" or "UFST" owned? | 17:51.03 |
| ray_laptop | Robin_Watts_: I don't think it is a patent issue -- I think it is just a LOT of work | 17:51.15 |
| mvrhel2 | UFST is a trademark of Monotype | 17:51.35 |
| Robin_Watts_ | Right, now I get it, thanks. | 17:51.46 |
| mvrhel2 | actually through the use of correlation methods, it could be possible to speed up the process of chopping into common parts | 17:52.34 |
| | that is automate is somewhat | 17:53.01 |
| | s/is/it/ | 17:53.06 |
| henrys | http://www.monotypeimaging.com/productsservices/Microtype_2.aspx | 17:53.14 |
| Robin_Watts_ | mvrhel2: Shouldn't be that hard, surely. It's something we'd only need to do once. | 17:53.19 |
| mvrhel2 | yes | 17:53.25 |
| ray_laptop | mvrhel2:right. Nowadays it might be possible to do a reasonably good approach by computer to find shareable parts | 17:53.31 |
| mvrhel2 | yes | 17:53.38 |
| Robin_Watts_ | henrys: According to that, they license the technology. (It's licensed to MPEG for example). | 17:54.13 |
| | We could find out how much a license would be... | 17:54.23 |
| ray_laptop | so we would start with the URW outline and try and "recompress" it | 17:54.26 |
| henrys | I really do wonder if that is the issue though - I believe the trademark font names are more important than the compression but I don't know for sure. | 17:54.32 |
| ray_laptop | henrys: when's the last time you looked to see if your printer said it had "real Monotyype Courier" ??? | 17:55.26 |
| Robin_Watts_ | In fact, they have released some stuff for use in OpenType fonts. | 17:55.41 |
| ray_laptop | the end user market just takes it for granted | 17:55.41 |
| Robin_Watts_ | http://www.w3.org/Submission/MTX/ | 17:57.07 |
| henrys | I look at the list of fonts supported by a printer before buying it, I assume that figures large in marketing folks thinking justified or otherwise. | 17:57.12 |
| | I guess I could ask norbert. | 17:58.05 |
| ray_laptop | I think trying to convince companies (at least PCL companies at the low end) to go with other than UFST is pushing a rope, but I _still_ think we need to understand the pricing issue -- monotype _has_ to be looking for increased revenue over just the fonts, otherwise, why bother | 18:03.42 |
| henrys | mvrhel2:can you ask if the pricing compared monotype pcl + monotype fonts vs. artfiex pcl + monotype fonts or our pcl + urw fonts? | 18:10.08 |
| | or maybe it is just best for miles to take over. | 18:11.51 |
| mvrhel2 | yes. I dont know what miles show them. MQ said miles had very high prices | 18:12.15 |
| henrys | I am not sure about monotype pcl but we know from other customers we are cheaper than adobe on PDF or PS.... oddd | 18:13.31 |
| | I'm going to talk to norbert about fonts. | 18:14.56 |
| Robin_Watts_ | Surely, as mvrhel2 says, when you're selling a PC based RIP the size of the fonts is irrelevant. It's probably the naming. | 18:15.51 |
| mvrhel2 | yes it made no snese to me | 18:15.54 |
| henrys | well I went ahead and asked. | 18:19.24 |
| mvrhel2 | my issue with the whole thing is that it is hard to place a price out there with respect to a ref system. their customer would really need to negotiate that with the pdl vendor | 18:26.15 |
| | it is possible that the prices that MI showed them did not include everything or were "subject to change" | 18:27.26 |
| | in the end, it would make sense for company C to offer their customers lots of choices | 18:28.05 |
| Robin_Watts_ | It's also possible that MI told them "we'll partner with you, but only if we're your sole partner". | 18:28.17 |
| mvrhel2 | there is that | 18:28.27 |
| henrys | all they need to say is 1000.00 laser for a reference system right? | 18:29.47 |
| Robin_Watts_ | Aargh. This ROP code is killing me. | 19:15.39 |
| henrys | Robin_Watts_:I don't think anyone since peter has been into it as deeply as you. | 19:17.33 |
| Robin_Watts_ | I can't blame peter for this. This is my own code. | 19:17.56 |
| | Well, I can blame him for the framework I have to fit into, but... | 19:18.10 |
| | I have a bug in one case of my code. | 19:18.25 |
| | I'm comparing pamcmyk4 to plank, and can see 2 classes of difference. | 19:18.48 |
| | One is my fault somewhere. | 19:18.54 |
| | The other one looks like a progression. | 19:19.00 |
| | Ah. Might have it. | 19:21.30 |
| | Aha! Exact match. | 19:23.14 |
| | I have a slight worry that all this special case "4 bit planar cmyk" code is bloating the build for people that will never select that device. | 19:24.14 |
| mvrhel2 | bbiaw | 20:29.51 |
| | whew. I think I finally have this halftone simple form stuff figured out | 22:27.39 |
| | ok it seems to be working too | 22:34.08 |
| | hmm maybe not | 22:34.45 |
| | argh. I will have to fool with this tonight. | 22:37.40 |
| | have to get ready for my son's birthday party | 22:37.55 |
| henrys | have fun mvrhel2! | 22:47.45 |
| mvrhel2 | bowling and video games. should be a fun time | 22:48.07 |
| sebras | wow. bug 692500 contains a LuaTeX-generated PDF that contains no less than 20200 shadings. no wonder it takes ages to draw... | 23:34.22 |
| | sometimes I wonder if it would be worthwhile to fix the PDF generators instead of trying to optimize PDF viewers... ;) | 23:35.18 |
| | Forward 1 day (to 2011/09/24)>>> | |
IRC Logs