stegfs-developing Mailing List for stegfs (Page 2)
Brought to you by:
albinoloverats,
sockeschuhregel
You can subscribe to this list here.
2003 |
Jan
|
Feb
|
Mar
(20) |
Apr
(8) |
May
|
Jun
(2) |
Jul
(3) |
Aug
|
Sep
(2) |
Oct
|
Nov
|
Dec
|
---|---|---|---|---|---|---|---|---|---|---|---|---|
2004 |
Jan
(3) |
Feb
|
Mar
(2) |
Apr
|
May
(17) |
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
2005 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
(1) |
Nov
|
Dec
|
From: Sebastian U. <seb...@cc...> - 2003-07-19 10:40:45
|
On Sat, 12 Jul 2003 12:49:06 +0200 (CEST) Joost Remijn <re...@ei...> wrote: > Hi there, Hi, > I'm Joost Remijn. I did look into porting stegfs to 2.4 myself some > time ago but it proved much harder then i thought. I'm not a good > kernel developer (yet) but i'm willing to help out. > > I have a job as a programmer at Eidetica (http://www.eidetica.com/) > and we make a search engine and other textmining related software. > This project however is purely a personal interest. Nearly everybody on this list had plans in the past to port stegfs to 2.4 :-) That's very interesting. > Oh yeah, i'm from the netherlands. Never mind, andreas and i are even from "old Europe" Germany ;-) > As a first question : > Why is there no source to be found in the sourceforge project? Or did > i overlook this somehow. We are working on the theoretic part in the moment. Feel free to take a look at the cvs on the project site and make some comments to the document if you have some ... Seems that all people are really busy in the summer and the project isn't going to proceed in the last weeks. But im sure that there will be some progress in the next months :-) Joost, would you like to become a stegfs developer for our project with the corresponding rights on our project site ? If you would like to, if will set it up. Just give me your sourceforge username ... -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Politik beginnt mit der Wahrnehmung von Wirklichkeit. -------------------------------------------------------- Dr. Guido Westerwelle (*1961), FDP-Parteivorsitzender, Rechtsanwalt |
From: Joost R. <re...@ei...> - 2003-07-12 10:49:24
|
Hi there, I'm Joost Remijn. I did look into porting stegfs to 2.4 myself some time ago but it proved much harder then i thought. I'm not a good kernel developer (yet) but i'm willing to help out. I have a job as a programmer at Eidetica (http://www.eidetica.com/) and we make a search engine and other textmining related software. This project however is purely a personal interest. Oh yeah, i'm from the netherlands. As a first question : Why is there no source to be found in the sourceforge project? Or did i overlook this somehow. Joost Remijn On Sat, 12 Jul 2003, Sebastian Urbach wrote: > Hi, > > I would like to welcome a new developer to this list :-) It would be > great, if you can tell us a little bit about yourself and your wishes > for stegfs and so on :-) > > -- > > Mit freundlichen Gruessen / yours sincerely > > Sebastian Urbach > > -------------------------------------------------------- > Politik beginnt mit der Wahrnehmung von Wirklichkeit. > -------------------------------------------------------- > Dr. Guido Westerwelle (*1961), > FDP-Parteivorsitzender, Rechtsanwalt > |
From: Sebastian U. <seb...@cc...> - 2003-07-12 10:37:09
|
Hi, I would like to welcome a new developer to this list :-) It would be great, if you can tell us a little bit about yourself and your wishes for stegfs and so on :-) -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Politik beginnt mit der Wahrnehmung von Wirklichkeit. -------------------------------------------------------- Dr. Guido Westerwelle (*1961), FDP-Parteivorsitzender, Rechtsanwalt |
From: Sebastian U. <seb...@cc...> - 2003-06-25 22:10:19
|
Hi all, There are some good news. Im running the 2.4.21 Kernel since yesterday and i found the kernel crypto api included in it ! I think that a very interesting fact because it should help us a lot in the near future :-) -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Mit oder ohne Religion koennen sich gute Menschen anstaendig verhalten und schlechte Menschen Boeses tun; doch damit gute Menschen Boeses tun, dafuer braucht es Religion -------------------------------------------------------- Steven Weinberg (*1933), Astrophysiker und Nobelpreistraeger |
From: Sebastian U. <seb...@cc...> - 2003-06-09 09:23:22
|
Hi all out there, Andreas and i are really busy in the moment, but we want to go on with stegfs developing when we have the possibility within the next weeks. -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Von all den Dingen die mir sind verloren gegangen, hab ich am meisten an meinem Verstand gehangen. -------------------------------------------------------- Ozzy Osbourne (*1948), eigtl. John Michael Osbourne. Musiker, Komponist und Schauspieler. |
From: Sebastian U. <seb...@cc...> - 2003-04-30 16:36:33
|
On 29 Apr 2003 08:29:08 -0400 Craig <cr...@ac...> wrote: > Hello all, > > What patch is currently being worked on. I didn't see any versions of > a patch within cvs. I would like to get started on getting the patch > to run on the newer kernel but want to keep in sync with everybody > else and their progress. I see that there is a task for the diff > between 2.4 and 2.6 completed. Is this for the stegfs patch or just > the diff or the kernel itself. Hi, The tasks are just things which we want to check before we gonna start the rewrite of stegfs. We want to offer the last stegfs version from Andrew D. McDonald with all available patches which are not inlcuded in his version till now. But it is clear that it wont work with actual kernel versions. That's one of the reasons why we gonna start the complete rewrite in a few weeks. -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Die Realitaetsverweigerung, der wir heute in Deutschland allenthalben begegnen, ist erneut die wichtigste Quelle unserer Irrtuemer, Illusionen und damit unvermeidlich kommender Fehlschlaege. -------------------------------------------------------- Prof. Dr. jur. Arnulf Baring (*1932), deutscher Zeithistoriker und Politologe |
From: Craig <cr...@ac...> - 2003-04-29 12:29:15
|
Hello all, What patch is currently being worked on. I didn't see any versions of a patch within cvs. I would like to get started on getting the patch to run on the newer kernel but want to keep in sync with everybody else and their progress. I see that there is a task for the diff between 2.4 and 2.6 completed. Is this for the stegfs patch or just the diff or the kernel itself. Thx all, Craig |
From: Sebastian U. <seb...@cc...> - 2003-04-28 20:51:26
|
Hi, I want to welcome another new developers to this list, it would be great if you can give us informations about your plans / ideas for stegfs and how you can help with developing from your point of view. Some personal informations would be also very helpful :-) It would be great if you are interested in becoming an active developer for the project :-) The next step would be that i give the you the details / permissions to do that ... Andrew, as i told you a few weeks ago, a development platform would be a great idea for stegfs and the community is getting bigger and bigger :-) -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Die Realitaetsverweigerung, der wir heute in Deutschland allenthalben begegnen, ist erneut die wichtigste Quelle unserer Irrtuemer, Illusionen und damit unvermeidlich kommender Fehlschlaege. -------------------------------------------------------- Prof. Dr. jur. Arnulf Baring (*1932), deutscher Zeithistoriker und Politologe |
From: Sebastian U. <seb...@cc...> - 2003-04-24 13:24:11
|
Hi, First i want to welcome two new developers to this list, it would be great if both of you can give us informations about your plans / ideas for stegfs and how you can help with developing from your point of view. Some personal informations would be also very helpful :-) It would be great if you both are interested in becoming active developers for the project :-) The next step would be that i give the you the details / permissions to do that ... John, are you working on some test software for Stegfs ? Last time you wrote that you want to finish the test software before we gonna start coding ... It would be very interesting if you can give us / me more information about the planned test software, something like what is tested exactly and so on :-) I think that the operating system port task is something for the future because we want to do the unix things first :-) -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Die Realitaetsverweigerung, der wir heute in Deutschland allenthalben begegnen, ist erneut die wichtigste Quelle unserer Irrtuemer, Illusionen und damit unvermeidlich kommender Fehlschlaege. -------------------------------------------------------- Prof. Dr. jur. Arnulf Baring (*1932), deutscher Zeithistoriker und Politologe |
From: Sebastian U. <seb...@cc...> - 2003-04-10 12:31:36
|
On Tue, 1 Apr 2003 20:15:52 +0100 john <jo...@po...> wrote: > 1. Because I think that Windows is an important target, I tried to find > out about writing file system drivers for Windows. It looks like the > only book on the subject is called "Windows NT File System Internals" > by Rajeev Nagar (published by OReilly). Unfortunately it is out of > print so I have not been able to get a copy. I had spoken to OReilly here in Germany today. The couldn't find the book in her own archive anymore, that means that the don't have even one book left ... The told me that the book isn't available anymore since January 2002. The are also not very optimistic that we can find the book on e-bay or something else :-( > 2. I originally volunteered to do testing for you. So I would need to > write some test programs (or to find some that someone else has > written). So that should be my first task really. I would like to have > the tests ready before anyone writes the code, so that we can test as > we go along. I have started the task "Test Software" today. The task is crying for input :-) > One of the tasks on Sourceforge is to try to compile the original > Andrew McDonald version on 2.4 or 2.5. Like I told you before, I tried > to do this once, but gave up. But at least I managed to make a > read-only version that works with 2.4. I will get that to work with 2.5. I put these informations in the task ... -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Die Realitaetsverweigerung, der wir heute in Deutschland allenthalben begegnen, ist erneut die wichtigste Quelle unserer Irrtuemer, Illusionen und damit unvermeidlich kommender Fehlschlaege. -------------------------------------------------------- Prof. Dr. jur. Arnulf Baring (*1932), deutscher Zeithistoriker und Politologe |
From: Sebastian U. <seb...@cc...> - 2003-04-02 12:48:22
|
Hi, Just want to say that Operting System Port Task is created with subproject Windows. -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Die Realitaetsverweigerung, der wir heute in Deutschland allenthalben begegnen, ist erneut die wichtigste Quelle unserer Irrtuemer, Illusionen und damit unvermeidlich kommender Fehlschlaege. -------------------------------------------------------- Prof. Dr. jur. Arnulf Baring (*1932), deutscher Zeithistoriker und Politologe |
From: Sebastian U. <seb...@cc...> - 2003-04-02 12:35:51
|
On Tue, 1 Apr 2003 20:15:52 +0100 john <jo...@po...> wrote: > On 2003.03.29 14:40 Sebastian Urbach wrote: > John, would be > > great if you can put some tasks on the site which should be done > > before we start to programming. > > I've really told you about most of my ideas already. > > I've just had 2 more thoughts: > > 1. Because I think that Windows is an important target, I tried to find > out about writing file system drivers for Windows. It looks like the > only book on the subject is called "Windows NT File System Internals" > by Rajeev Nagar (published by OReilly). Unfortunately it is out of > print so I have not been able to get a copy. > > 2. I originally volunteered to do testing for you. So I would need to > write some test programs (or to find some that someone else has > written). So that should be my first task really. I would like to have > the tests ready before anyone writes the code, so that we can test as > we go along. > > One of the tasks on Sourceforge is to try to compile the original > Andrew McDonald version on 2.4 or 2.5. Like I told you before, I tried > to do this once, but gave up. But at least I managed to make a > read-only version that works with 2.4. I will get that to work with 2.5. I understand that :-) Andreas and i just think that it would be a good idea to start tasks on the sourceforge site, which are good if you have something to investigat or checked from the developers point of view, before we start writing the code. An example : Andreas had an idea a few days ago to use orthogonal codes for StegFS, nobody nows if that can work with StegFS. So he started a task and is going to catch up on the Orthogonal stuff and put it in the task. He set a deadline for this task and if the time is gone we well se if there is a chance to do it. Everbody is invited to add informations to the tasks, create new tasks and so on, but nobody is forced to do that. I will add your Information concerning the kernel and compiling things within the next days if you don't do it before :-) I think i will open up a operating system port task with a windows section inside. I would be happy if you put some Informations in it :-) -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Die Realitaetsverweigerung, der wir heute in Deutschland allenthalben begegnen, ist erneut die wichtigste Quelle unserer Irrtuemer, Illusionen und damit unvermeidlich kommender Fehlschlaege. -------------------------------------------------------- Prof. Dr. jur. Arnulf Baring (*1932), deutscher Zeithistoriker und Politologe |
From: john <jo...@po...> - 2003-04-01 19:16:32
|
On 2003.03.29 14:40 Sebastian Urbach wrote: John, would be > great if you can put some tasks on the site which should be done > before we start to programming. I've really told you about most of my ideas already. I've just had 2 more thoughts: 1. Because I think that Windows is an important target, I tried to find=20 out about writing file system drivers for Windows. It looks like the=20 only book on the subject is called "Windows NT File System Internals"=20 by Rajeev Nagar (published by OReilly). Unfortunately it is out of=20 print so I have not been able to get a copy. 2. I originally volunteered to do testing for you. So I would need to=20 write some test programs (or to find some that someone else has=20 written). So that should be my first task really. I would like to have=20 the tests ready before anyone writes the code, so that we can test as=20 we go along. One of the tasks on Sourceforge is to try to compile the original=20 Andrew McDonald version on 2.4 or 2.5. Like I told you before, I tried=20 to do this once, but gave up. But at least I managed to make a=20 read-only version that works with 2.4. I will get that to work with 2.5. John |
From: Andrew M. <an...@mc...> - 2003-03-31 23:24:10
|
On Sat, Mar 22, 2003 at 12:45:01PM +0100, Sebastian Urbach wrote: >=20 > Andrew, are you alive ? Did you received the last mails over this > list ? We are waiting for starting the structure discussion till you > say something ... I've just got back from a week's holiday and am slowly catching up on several hundred e-mails (far too many mailing lists....). Don't wait for me if you want to get on with any more detailed discussion. (I'll try to give some more comments tomorrow, maybe). Regards, Andrew --=20 Andrew McDonald E-mail: an...@mc... http://www.mcdonald.org.uk/andrew/ |
From: Sebastian U. <seb...@cc...> - 2003-03-29 14:51:37
|
Hi, seems that Andrew doesn't have enough time to answer. Meanwhile we should start to go on with the basic research. Andreas and i started to put some Tracks on the project site. It would be great if everbody can read them, choose a track or more to work on :-) John, would be great if you can put some tasks on the site which should be done before we start to programming. From my point of view the tracks are more specific theortic work before we start the real programming :-) Andrew, it would be great if you answer the mails from the last weeks when you have enoug time, im sure that we can integrate you every time in our work ... 'a nice weekend for all of you. -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Die Realitaetsverweigerung, der wir heute in Deutschland allenthalben begegnen, ist erneut die wichtigste Quelle unserer Irrtuemer, Illusionen und damit unvermeidlich kommender Fehlschlaege. -------------------------------------------------------- Prof. Dr. jur. Arnulf Baring (*1932), deutscher Zeithistoriker und Politologe |
From: Sebastian U. <seb...@cc...> - 2003-03-22 11:55:37
|
Hi all, Andrew, are you alive ? Did you received the last mails over this list ? We are waiting for starting the structure discussion till you say something ... -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Die Realitaetsverweigerung, der wir heute in Deutschland allenthalben begegnen, ist erneut die wichtigste Quelle unserer Irrtuemer, Illusionen und damit unvermeidlich kommender Fehlschlaege. -------------------------------------------------------- Prof. Dr. jur. Arnulf Baring (*1932), deutscher Zeithistoriker und Politologe |
From: Sebastian U. <seb...@cc...> - 2003-03-19 12:32:56
|
Hi, Andrew, could you please give us a signal when you are finished with reading our paper from the cvs ? If you have any ideas / requirements than please open your own section in this paper and write the ideas down. Another question to you is, if you want to be a developer for this project, or if you "just" want to be a member of this mailing list ? If you want to be an active developer, you should have an sourceforge user account and i would need your username to give you the developer privileges. We would be definitively very happy if you want to become a developer for this project :-) The other thing i want to talk about is, that we should start the structure discussion for stegfs if everybody is finished with reading the paper and writing ideas down. Performance discussion is good, but i think what we need first is to discuss the basic things ... -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Die Realitaetsverweigerung, der wir heute in Deutschland allenthalben begegnen, ist erneut die wichtigste Quelle unserer Irrtuemer, Illusionen und damit unvermeidlich kommender Fehlschlaege. -------------------------------------------------------- Prof. Dr. jur. Arnulf Baring (*1932), deutscher Zeithistoriker und Politologe |
From: Sebastian U. <seb...@cc...> - 2003-03-18 17:56:22
|
Hi, Andrew, could you please send me all available patches for the last stegFS Version 1.1.4 ? I will test the patches and load an patched Version up on the source forge page and link it from our Website. The other way is that you do it yourself and load it up to your own site :-) So, what do you prefer ? -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Die Realitaetsverweigerung, der wir heute in Deutschland allenthalben begegnen, ist erneut die wichtigste Quelle unserer Irrtuemer, Illusionen und damit unvermeidlich kommender Fehlschlaege. -------------------------------------------------------- Prof. Dr. jur. Arnulf Baring (*1932), deutscher Zeithistoriker und Politologe |
From: SockeSchuhRegel <soc...@so...> - 2003-03-18 10:45:31
|
Hi everyone! I'm Andreas Petter (the only developer that doesn't get his mailsystem running a s he wants) and finally tricked lists.sourceforge.net to accept my emails, too. However since, reception of emails waorked fine I excitedly listend to the block size debate. I think that it is not elegant, since we would have to check sever al different cases and I'm not sure if this is possible, but there seems to be a slight but rather dif ficult chance :)... Since standard block size is somewhere around 2048 or 4096 bytes and files norma lly do not end at multiples of 4096 we could save those checksums (of lots of bl ocks which fit into the rest of the last block into this block thereby saving sp ace and write operations. Hm. If the standard file ends (in the middle) at 2048 bytes of the last block, we got space for (2048/(5*128)) block checksums = 2 * 4 096 + 2048 bytes = 10240 per file (and even more if we think about saving the checksum of the last block of the file only once per written block). If this does not suffice we just could use another block. I'm currently thinking about if we could use orthogonal codes for even better da ta integrity, using CDM, code divisions multiplex. To apply multiplexing techniq ues we could think about the disk as not being space but a linear time space, wh ich has to be divided between several users, starting with time 0 at block or by te 0 and ending at the last block of the disc (time X). What has currently been done was time division multiplex, with repeated sending because of collisions. S aving using a spread spectrum technique, which first spreads the bits and then m ultiply by orthogonal codes could prevent overwriting while still keeping the se curity requirements more than it can be done as proposed by PTZ (who in fact man aged to implement theoretical defined plausible deniability while not overwritin g and not repilcating blocks) (that it is more stegano-like and therefore more s ecure is the opinion of the author, only :) ). Within this scheme it could be ve ry useful to use error correction codes instead of error detection codes, thereby reducing the probabilit y of getting overwritten bits enormously. However, I have no plan how to calcula te the probability (since more blocks are spent (they may and should be ovverwri tten, however) and loss may not be defined as easy as it is for blocks which jus t get overwritten) and we would have to discuss that deeper and think about that longer to get a working scheme. Sorry for the long mail, Greetings, -Andreas |
From: Andrew M. <an...@mc...> - 2003-03-17 19:52:54
|
On Mon, Mar 17, 2003 at 06:50:47PM +0000, john wrote: > On 2003.03.16 20:46 Andrew McDonald wrote: > > >If the verification data held in the block table file could be moved > >into the disk blocks themselves then some performance improvment might > >be obtained from this. (i.e. we don't need to access a separate bit of > >the disk for every block accessed). > > If we do this sort of thing, then I suppose each disk block would > contain something like 1920 bytes of data and 128 bytes of verification > data (I'm assuming 2K blocks; the btab entries are currently 128 > bytes). This means that the amount of data in each block is not a power > of 2. Yes. This is the key problem, and why it was always on the "to think about later" list. -- Andrew McDonald E-mail: an...@mc... http://www.mcdonald.org.uk/andrew/ |
From: john <jo...@po...> - 2003-03-17 19:50:42
|
On 2003.03.17 18:50 john wrote: > On the other hand there is a project called "cdfs" that *does* allow=20 > you to mount a video CD under Linux. It might be worth looking at=20 > this to see whether they have had to do any special workarounds=20 > because of the strange sector size. OK, I've taken my own advice and had a look at the cdfs code. I *think*=20 that the following is accurate, but I am not a kernel hacker and I may=20 be wrong. When you open a file, you return to the kernel an inode structure which=20 contains a pointer to some "file_operations" functions and a pointer to=20 some "address_operations" functions. You have 2 choices: 1. Implement all the file_operations yourself. This is hard work and=20 you lose all the benefits that VFS gives you. 2. Use the generic functions that VFS supplies for the file_operations.=20 These will do all the hard work of managing file pointers, buffering=20 and so on. Then all you have to do is to implement the readpage and=20 writepage methods in address_operations, which read and write the file=20 a sector at a time. The generic file_operations assume that the sector size is a power of 2=20 (1024, 2048, 4096,...) so if you make choice (2) you have to provide=20 methods that will read or write a chunk of a file which is of such a=20 size, and which is aligned on a sector boundary. Because the video CD sectors are not of the right size, the cdfs code=20 deals with this by splitting up the request to read a 2048-byte-aligned=20 sector into 2 reads: the first takes some of the data from part of a=20 2352-byte sector, and the second takes the rest of the data from the=20 start of the next sector. Since 99% of reads on a video file are going=20 to be sequential, this need not be too inefficient. But if we are going=20 to use this idea for random reads on arbitrary files, we are going to=20 take a big performance hit because each 'logical' sector read by VFS is=20 going to have to be implemented as 2 'physical' sector reads. So I would suggest that we *don't* put the verification data into the=20 disk blocks, so that we don't need to use non-standard sector sizes,=20 because it will be inefficient. This means that we will have to retain=20 the 'btab' file. Of course, with the 'btab' file we still have to do 2 physical writes=20 per logical write: one to the sector we are updating, and the other to=20 the btab file. But the btab file is likely to be fairly small, so it is=20 likely to be buffered in memory for most of the time. John |
From: john <jo...@po...> - 2003-03-17 18:51:23
|
On 2003.03.16 20:46 Andrew McDonald wrote: > If the verification data held in the block table file could be moved > into the disk blocks themselves then some performance improvment might > be obtained from this. (i.e. we don't need to access a separate bit of > the disk for every block accessed). If we do this sort of thing, then I suppose each disk block would=20 contain something like 1920 bytes of data and 128 bytes of verification=20 data (I'm assuming 2K blocks; the btab entries are currently 128=20 bytes). This means that the amount of data in each block is not a power=20 of 2. I seem to remember reading that the linux VFS does not like this sort=20 of thing. For example, under Windows you can mount a "video CD" and=20 read it like a regular file, but under Linux you have to use ioctls to=20 read the data. A video CD has 2352-byte sectors. On the other hand there is a project called "cdfs" that *does* allow=20 you to mount a video CD under Linux. It might be worth looking at this=20 to see whether they have had to do any special workarounds because of=20 the strange sector size. John |
From: Sebastian U. <seb...@cc...> - 2003-03-17 14:17:43
|
Hi, Just want to say that we are 4 people on this list since Andrew is on the road :-) Andreas, John, Andrew and i (Sebastian). Andreas is still having some problems with his mailsystem and he is on the CEBIT today. I hope that he is able to start the discussion when he is back tonight. -- Mit freundlichen Gruessen / yours sincerely Sebastian Urbach -------------------------------------------------------- Die Realitaetsverweigerung, der wir heute in Deutschland allenthalben begegnen, ist erneut die wichtigste Quelle unserer Irrtuemer, Illusionen und damit unvermeidlich kommender Fehlschlaege. -------------------------------------------------------- Prof. Dr. jur. Arnulf Baring (*1932), deutscher Zeithistoriker und Politologe |
From: Andrew M. <an...@mc...> - 2003-03-16 21:01:07
|
On Sun, Mar 16, 2003 at 08:03:57PM +0000, john wrote: >=20 > But I think we get that automatically. Under Linux you can put any file= =20 > system you like on a CDROM; it doesn't have to be the ISO file system.=20 > The ext2 filesystem works just fine. You just make a 700MB file, use=20 > mke2fs to make a filesystem in it, mount it with the loopback device,=20 > and copy whatever you want into it. Then you write the whole 700M file=20 > onto the CDROM with cdrecord just as though it were an ISO image. If=20 > this works with ext2 then it is bound to work with stegfs. The only=20 > slight complication is the extra "btab" file that stegfs needs, and I=20 > have proposed that this should live inside the filesystem instead of=20 > being separate. I agree. I've got a feeling that you can move the btab file onto the filesystem after you've created it (so it is in the "visible" ext2 part). But, it is a while since I've played with this. The file access for the btab file from within the kernel has always felt a bit hairy. The current code is based on the khttpd. I've got a feeling that there are some potential problems doing it this way where you haven't got a pid (as is the case when you're doing it from within the kernel). ext3 might well be a useful guide on how to do this properly (which didn't exist when I originally wrote stegfs). Another thing we orginally discussed, but I never got round to implementing was moving the block replication to a daemon. The replication is one of the real performance killers in StegFS. The idea here would be to only write the first copy immediately, a kernel daemon would then perform the replication later. (Again, kjournald might provide some sort of model to follow). Such a daemon could also do other things such as the filling of deleted blocks with random data (another slow task), verifying the integrity of the file system (and any required rereplication), and also some random change of unused blocks to disguise which blocks are really being used. Looking back at my original dissertation (which you can find at http://www.mcdonald.org.uk/StegFS/stegfs-dis.ps.gz), other things noted include the possibility of readahead/caching of decrypted blocks and the caching of the checksum verification on blocks. However, my tests at the time showed that we were disk bound rather than CPU bound. If the verification data held in the block table file could be moved into the disk blocks themselves then some performance improvment might be obtained from this. (i.e. we don't need to access a separate bit of the disk for every block accessed). Hopefully those initial thoughts are of some use. :-) --=20 Andrew McDonald E-mail: an...@mc... http://www.mcdonald.org.uk/andrew/ |
From: john <jo...@po...> - 2003-03-16 20:04:38
|
On 2003.03.16 14:01 Sebastian Urbach wrote: > Hi * :-) >=20 > There are some great news for us. Andrew contacted me and told me that > he put a link on his website www.mcdonald.org.uk/stegfs which points > right to stegfs.sourceforge.net (our project website) :-) >=20 > He also told me that he want to subscribe on our mailinglist and it is > done :-) I hope that he want to help us as far as he can and im > absolutely sure that Andrew could be a big help for us if he wants to > :-) >=20 > Let's se if he want to become a sourceforge developer for our project, > i hope that we together can convice hm :-) He also gave me a hint for > the diagramm stuff from John, well let's see if i can implement it :-) >=20 > Andrew, welcome on this list ! We are happy to have you with us :-) Andrew, welcome from me too. > PS: A friend of me posted a feature request on the source forge > project site, let's read it and think about it ... The request is for the ability to store encrypted data on a CDROM or=20 other read-only media. But I think we get that automatically. Under Linux you can put any file=20 system you like on a CDROM; it doesn't have to be the ISO file system.=20 The ext2 filesystem works just fine. You just make a 700MB file, use=20 mke2fs to make a filesystem in it, mount it with the loopback device,=20 and copy whatever you want into it. Then you write the whole 700M file=20 onto the CDROM with cdrecord just as though it were an ISO image. If=20 this works with ext2 then it is bound to work with stegfs. The only=20 slight complication is the extra "btab" file that stegfs needs, and I=20 have proposed that this should live inside the filesystem instead of=20 being separate. We might need to do a little extra work to make this work with Windows,=20 which only understands the ISO filesystem on CDROMs, but it can't be=20 very hard. John |