On Wed, Oct 25, 2000 at 03:46:27PM -0400, Havoc Pennington wrote: > > Computer writer Evan Leibovitch recently contacted the steering > committee to ask about the relationship of AbiWord to > OpenOffice. According to him, some members of the AbiWord team are > worried that Sun's code has been declared the canonical office suite > and that AbiWord will be entirely dropped, without consulting > them. Evan is writing some sort of article about this; but the views > he cited in mail to me are in my opinion an inaccurate description of > reality. It is true that such worries have been expressed both in public and in private, by members of the AbiWord development team. While this may not have been the intended message from the statements made upon the announcement of the GNOME foundation, some of those statements did come off in a way that might suggest this. Since then, many of these misconceptions have been cleared up. > > The actual situation to my knowledge is that all technical decisions > were put off until we had an elected foundation board and access to > the OpenOffice code. Our announcement at LinuxWorld was the > common-sense fact that we're going to use all the best code that's > available to us, and create a GNOME Office suite out of > that. OpenOffice is one such piece of code. This is not intended to > imply that any particular pieces of code will or won't be used in the > final product. > The GNOME Foundation has turned out better than I had originally imagined was possible. It turned out not to be an exclusive group, with admission by invitation only, as I had originally imagined. In fact, 5 AbiWord coders, including myself, are now members of the foundation, and we will be able to elect the board members from among our own, one hacker, one vote. In fact, this is one more vote per GNOME hacker than each of the large companies signed on to the GNOME foundation receive for the board. This is greatly reassuring for me, and suggests that GNOME will continue to be run in a fashion that will produce the best technical decisions, and a superior desktop environment for its users. It is also reassuring that decisions about the code will be made on a technical basis, and by our democratically elected leaders. (I still hope I can vote for you, Dom :-) > In a mail to abiword-dev shortly after the LinuxWorld announcement, I > tried to make clear that no decision in this area has been made, and > indeed hardly could be made until the Oct. 13 code release which > allows us to evaluate the Sun code. This is still true. No decision > has been made. Unfortunately I think my mail to abiword-dev was held > for moderation and judging by the archives, the moderator never > approved it; but the mail was partially quoted in followups IIRC. > > I'd like to take the opportunity to 1) apologize to the AbiWord team > for the confusion and 2) ask the AbiWord team to please participate in > discussions about this, and feel free to initiate such discussions. At > the same time, Sun should be actively working to meet and work with > existing GNOME Office hackers. The burden is on _both_ the AbiWord > and OpenOffice projects to communicate. I believe that the whole AbiWord community would agree with me in saying that you, Havoc, owe us no apology. I also believe that we would all agree that further dialogue with both the Open/StarOffice people, and the GNOME people, can only benifit all of our software. However, I know that several AbiWord people, including myself have posted to the OpenOffice lists, to little effect, and AFAIK, no OO/SO people have joined our lists. I point this out not to criticize the hackers at OO, but to suggest that we are already interested in communication. > My personal view is that having two office suites, or two office suite > components, with exactly the same design goals (full-featured end user > word processor) is entirely stupid. Therefore, we should have only one > project, and I think that _eventually_ _after all the technical issues > have been discussed_ and _the best course is evident_ the GNOME > Project should pick a single application of each type to go in the > suite called GNOME Office. > It would be nice if there didn't have to be competing Free Software projects. And I, and I suspect others, are willing to work toward integration. However, there are serious barriers to integration, as I will discuss below. > In my ideal world, the single resulting project would include both the > historical StarOffice team and the historical AbiWord team, and the > best code from all of StarOffice and AbiWord and Gnumeric and Guppi > and so on. The only reason that won't happen is ego or lack of > trying. So, I'm asking both teams to please try; and to think about > the end goal of delivering a full-featured office suite to end users > as soon as possible, putting ego aside. I disagree that the only reason for failure of integration is lack of will or effort, or an excess of ego. There are currently definite technical reasons, in addition to political reasons. > > If people don't make a serious and sustained effort to do this, it's > going to damage the prospects of free software office solutions. So I > hope people will take the task seriously. If the merger doesn't > happen, that's not so bad; but if people didn't honestly try there > will be lasting bitterness because the split will be gratuitous and > political in nature. If people really try and there's really a good > technical reason to have two projects, then we can get friendly > competition instead. > > I recognize that the barriers are huge. Some of the underlying > technologies, such as Bonobo and libgnomeui, are themselves immature; > the amount of code involved in OpenOffice and in GNOME Office is > enormous; there are multiple programming languages involved; there are > half a dozen target platforms; there are probably 150 hackers involved > in half a dozen different communities on different mailing lists. This > is not easy. > The major obstacles to integration, as I see them (in rough order of importance): (1) AbiWord is Cross-Platform This is the first thing we mention about AbiWord. This is the point I like to brag about most, since we are the most portable word processor around. AbiWord was designed internally to be XP from the ground up. [1] We are the only word processor on QNX. We run on AIX, IRIX64 and other big iron operating systems. We run on every 32 bit version of Windows. We run on every linux distribution, and on all the BSDs. Currently, SO runs on Solaris, Linux and Win32. While there continutes to be this discrepancy, AbiWord will not want to be absorbed into a less portable project. I don't see this as an easily resolved issue, especially since it also involves the further ease of porting to other unsupported platforms, which is currently not difficult for AbiWord. (2) The Attitude of Sun I feel bad writing this section, as it is highly critical, in some ways, of Sun. However, I feel it has to be said. The postition you stated in you email was "may the best code win." In other words, the direction of the merge should be dictated by the technical situation. However, I suspect that this is not the position Sun is planning to take. They have Freed StarOffice with the intent of it becoming the primary office suite for Linux (and maybe Win32 also). They seem to have a business plan, revolving around their upcoming StarPortal, that depends on this. Therefore, I feel that it is unlikely that they would take kindly to the suggestion that OpenOffice would abandoned in favor of AbiWord. In fact, I am virtually certain that regardless of the decision of the GNOME Foundation board, OpenOffice will continue on its present course. Thus, the possibility for decision on this issue seems slightly skewed, at least from this end of the table. As an example of the problems with Sun's attitude toward their cooperation with the larger community comes from their XML file format, an area that AbiWord folks will know is near and dear to my heart. When Sun first annouced their plans for OpenOffice, these plans included developing an XML file format, which they wanted to make into a standard for the free software community. I found this possibility exciting, and posted a message to the discussion mailing list at openoffice.org. I offered to work with the OpenOffice developers on developing this format. I supposed that cooperation between KWord, AbiWord and OpenOffice would allow us all to have a common format, something practically unheard-of in the software industry. I recieved no substantive responses, and certainly none from the Sun engineers who were creating the file format. Thus, on Oct. 13, when Sun released all of the code, they also released a 300 page file format spec, the basic foundations of which were therefore already laid, with no input from the community that Sun wanted to adopt this format. Instead of taking advantage of offers from the community, and creating goodwill and quality working realationships, Sun chose to go their own way. But most sad is their current position on the file format. They state [2] that "community control" is a desired goal, but a neccessary requirement is "OpenOffice must be the reference implementation for this file format." I don't think I would feel comfortable working with Sun to develop a file format that I knew they had veto power over. This demand also makes true standardization, through a body such as the W3C or OASIS essentially impossible. It seems that, in the realm of file formats, Sun is looking for a standard like they have created with Java, rather than the kind XML provides. As an example of how a company that really "gets it" work, I offer the following <quote src="BeOpen Interview"> BeOpen: Still, there are multiple journaling file systems -- XFS, JFS by IBM, ext3 and Reiserfs. Does SGI see an advantage if XFS is accepted as the standard journaling file system? Vrolyk: That's not the goal in Linux. You have to be very careful not to apply old proprietary modes of thought to what is essentially a completely new market with completely new rules. In the Open Source market there's no value to having it become the standard. The only value to SGI is that we make sure that Linux has the best technical solution. We don't care where it comes from, whether SGI or IBM or one of the Open Source community projects. What we do care about is that whatever gets implemented is as good as or better than XFS. We want to avoid fragmentation. If the Open Source community says, "Great, XFS is exactly what we need," we'll support it. If not, we'll kill it. We won't support competing standards. </quote> [3] If this was Sun's attitude toward StarOffice, I would most likely feel differently. (3) Internationalization I don't know much about the support for various languages in SO/OO, but in AbiWord everthing is Unicode internally, and we currently handle all single-byte languages, with multiple-byte support likely coming soon. We are internationalized to languages from Russian to Catalan, and would like to continue to maintain our commitment to all of our users, regardless of language. (4) License As Martin pointed out, SO/OO is dual licensed. Therefore, in order for AbiWord code to be used in SO, we would have to track down all of our contributors and ask their permission for their code to be used under the SISSL [4]. As we have more than 100 people on the list of contributors [5], this might be difficult, leaving aside the problems of getting them to agree to this change. (5) Sheer Size Given the huge size of the StarOffice code base, and the not insignificant size of the AbiWord source, and given their radically different design philosophies and strategies, merging the projects would present significant logistical difficulties. That's about it for obstacles that immediately spring to mind. > If people wanted to just jump in and really attack this problem, one > possible set of stuff to do might be this: > > - let's use gnome-office-list gnome org for now to discuss these issues > and plan future direction. Hopefully everyone who got the original message is subscribed. > > - a short-term goal should be to get all the office code on a single > CVS server and with a single mailing list, etc. framework. The two > candidates here are openoffice.org and gnome.org. If it's done on > gnome.org, openoffice.org should basically shut down its independent > hosting. Though purely from a load-balancing standpoint, using > their setup is sensible. To start, the GNOME code and Sun code > can be in separate modules. This would be a truly massive undertaking, coordinating the integration of sources from three different locations. However, as SourceGear plays a less and less active role in AbiWord development, this might be a move they would welcome. However, I think this is a step best put off until after the next thing you describe. > > - If possible, I would like to see a short-term release of the Sun > codebase which basically patches up breakage caused by removing > proprietary stuff, makes dialogs appear in real toplevel windows, > and nothing else. The reason is that the full-featured fully-GNOME-aware > office suite is going to take a couple years, and it would be nice > to have a relatively complete suite users could use in the meantime. > > Going ahead and finishing AbiWord 1.0 before introducing major > breakage wouldn't be a bad idea either. > > Hackers tend to have the "if it's not perfect it's useless" > attitude which I think is broken; learning from Mozilla, an interim > release that saved us from 2 extra years of closed-source unfixable > Netscape 4.x would have been invaluable. > > So IMHO the discussion should perhaps focus on plans for _after_ > the next stable release of each project. I agree with this whole-heartedly. AbiWord is definitely nearing 0.9.0, our goal for 1.0 feature-completeness. We are essentially missing only one major backend feature, and a few on the front end. Once we get AbiWord 0.9 stabilized and we beging to move towards 1.0 (our first stable release) we will be able to think about major project changes. I suspect the same is true of the StarOffice people. > > - Modulo that short-term release, I would suggest creating a new source > tree in the agreed-upon shared CVS repository. There I would create > a new package; and then in a planned way, with all the hacker teams > cooperating, start moving chunks of code into the new GNOME Office. Before this can happen, I think the important issues that I and others have mentioned need to be addressed satifactorily, as I hope they can be. > > - Another thing we can learn from Mozilla; often the only way to get > people to decide between two chunks of code is to write a new chunk > of code that replaces both chunks with something better. Though > obviously it would be better to avoid rewriting something like 7 > million lines of code, judicious use of this tactic for portions of > the office suite can't hurt. It gives all the hackers an investment and > interest in the new code. I agree that in a merge substantial rewrites will have to take place, but I see them as less likely to happen in AbiWord as in StarOffice (if only because we have so much less code). > > - Recognize that a real office suite with native look-and-feel that > can genuinely compete with Microsoft's is a couple years away. You > may as well go slow and get it right. I think this is the most important lesson from Mozilla - it takes time. AbiWord has gotten to where we are now from nothing in two years, and in another two plenty could happen. Great code doesn't appear overnight, and whatever solution we come up with will take lots of work. > > - MOST IMPORTANTLY: keep it technical. As soon as anyone escalates to > fear tactics and conspiracy theories then nothing productive will > get done. Ignore the flames; don't do conspiracy theories; get > together and solve the problem at hand, which is to deliver a > full-featured open source competitor to MS Office sometime in the > next couple years. (If people don't agree that's the problem, > then maybe start by saying what problem you think the project > should be addressing.) I feel confident that this is achieveable. I'm not so confident that it will happen as easily as you predict. People are either going to have to make significant compromises, or not be part of the unified suite. It may end up better for AbiWord or StarOffice to remain as a competitor to the other, although hopefully that won't be neccessary. I hope we can continute to discuss this civilly. Thanks, Havoc, for starting this discussion. [1] http://www.abisource.com/talks/lwce2000ny/sld001.phtml [2] http://xml.openoffice.org/ [3] http://beopen.com/stories/article.php?id=24&page=2 [4] http://www.openoffice.org/licenses/sissl_license.html [5] http://www.abisource.com/lxr/source/abi/CREDITS.TXT sam th sam uchicago edu http://www.abisource.com/~sam/ GnuPG Key: http://pgp5.ai.mit.edu:11371/pks/lookup?op=get&search=0xCABD33FC
Attachment:
binfyirYNhKAv.bin
Description: PGP Key 0xCABD33FC.