Forum OpenACS Q&A: OpenNSD - should we or shouldn't we?
a few weeks ago. I said then that I would start a thread about it,
and here it is.
There were a number of factors which doomed the previous effort. The
main ones were lack of time on the part of the organizers, and the
upcoming 4.0 release. Basically, we appointed gatekeepers to review
patches, much as is done for OpenACS, but all of the gatekeepers (I
was one) were too busy to do the reviewing in a timely manner. And
many of us had a feeling that perhaps we were wasting our time
enhancing 3.x, since we thought 4.0 would be out Real Soon Now. That
was about a year and a half ago...
With 4.0 still impending, and looking more likely now that Tcl 8.4 has
been released, I don't think it makes much sense for any of us to
spend time enhancing 3.x at this point. If we do anything at all, I
think we should just take the 3.3+ad13 and 3.4 tarballs, change the
identification string to say OpenNSD, tar them up with an appropriate
name, and make them available from a one-page site which just explains
that this is the current version and that there will be more of a
project structure and participation when 4.0 comes out. Anything more
will be too much of a distraction from the OpenACS and dotLRN
releases, and probably not worth the effort at this time.
If folks agree that this is the way to go, I'll set up the site and
make the tarballs. I'll give interested parties a chance to comment
on the text before it goes live. We can offer RPMs if someone else
wants to make them; that's not a project I can tackle right now.
Assuming folks want to go ahead with this there is one other thing to
decide, and that's what URL to use. opennsd.org/net are owned by "DRS
Expired Domain", which I presume is a cyber-squatter who snaps up
expired domains. Through register.com I could make an offer to buy
them for a minimum of $200 each, but I think what these folks do is
really slimey and I hate to give them the money. opennsd.com is owned
by Scott Goodwin, who would probably let us use it, but a .com address
really isn't right for a project like this.
We could reserve open-nsd in any variations we like -
org/net/com/info? They are all available. So how do folks feel about
going with the hyphenated domain name? It's not optimal, but
personaly I'd rather do that and get the right extension. I'm willing
to pay for it, though anyone who wants to chip in is welcome.
Let the discussion begin!
I really think that open-nsd is fine. People who are looking for webservers won't know opennsd or open-nsd (and probably not even aolserver).
Putting the tarballs up is good. I really think we need to get this project back on track. I really doubt whether there will be a 4.0 release soon anyway.
What about patches. I know that there are a lot of useful ones scattered about and without them it wouldn't do much good. I can send you my patch if you need it.
Definitly lets get some momentum on this project. Even if 4.0 comes out I am not sure it will allow all we want/need (i.e. namespaces, use of standard tcl libs & etc.).
I have a little time if you need some help coordinating this.
I would prefer to stay with some variation on OpenNSD - the name is already out there, and it might get confusing if we relaunch *and* rename the project.
From a development project view, I've heard the AOLserver folks have been officially told not to work with the community anymore because we've wasted their time trying to convince them we can help rather than hinder their process. I would hate to fork violently so that they decide to withdraw further into closing their license (although their prior code is dual-licensed under the GPL). That's a worst case scenario, and probably unlikely.
Still, their code is getting released, and it is damn good. Would hate to risk any of that at all.
Simply in the interest of getting a better sense of why this is a good idea, can a short history of the problem be written, or at least can a direction to a previous summary be given? As well as summary of what's happened since the split and why we should follow through.
If we do split, I think that we should make it a subproject of dev.openacs.org, which is where the next gen site is right now.
I would want to know the source of this before taking it too seriously. Part of the reason for starting the OpenNSD project in the first place was because the AOL project had a person who was, shall we say, "not temperamentally well-suited to dealing with the public" in the role of being the public contact person. He was generally bad-tempered and dissed a lot of people's honest efforts to contribute. He wouldn't even take Rob Mayoff's patches, which is why we have aD releases of AOLserver. So if it was him who was complaining about the community, you'd better have the world's largest grain of salt handy.
There are (at least) two motivations for creating a separate project:
<li>The name AOLserver is a liability. We've only run into it peripherally, but it is out there. A business type recently pointed out to me that when you go in to try to sell to a decent-sized corporation, you are probably talking with someone who reads the business section, knows all about the problems at AOLTimeWarner, and probably has even lost money personally on AOL stock. These folks have a reflexive, knee-jerk negative reaction to anything having to do with AOL.
<li>The community is going to want to make changes to nsd that AOL will not want in their version. They quite rightly have to be somewhat conservative about what changes they accept (they should just be nicer about it, IMHO). Anything that doesn't have direct value to AOL should not be included as it's not worth the potential source of new bugs.
I'm thinking that what we might want to do, and this gets back to Jon's question too, is to provide the "base" tarballs, which would be 3.3+ad13 and 3.4 as they exist today (and 4.0 when it's ready) and then publish all our changes in the form of patches. People can pick and choose their patches, or download a "patched tarball" that will give them the latest OpenNSD. That way we don't fork, exactly, but add on to what AOL has released. It could get "interesting" when several people change the same area, though; patches would have to work both when applied alone and in the fully-patched version.
We definitely don't have to worry about upsetting the AOL folks who matter; Jim Davidson was all in favor of the OpenNSD project last time around.
I disagree that the OpenNSD site should be at openacs.org, though. There are plenty of folks out there who use nsd but don't use OpenACS; they have important things to contribute too. They are more likely to participate if OpenNSD is a free-standing project.
Perhaps OpenACS.org should have an NS.openACS.org site that keeps a current AOLserver with a customized configuration complete with OpenACS related patches?
In other words, create a sort of evolving ad13+.. project for OpenACS.org
This arrangement hopefully:
1. respects the aolserver community and their work (per Talli's comments)
2. keeps the OpenACS community from having to duplicate much of the AOLserver work as the software evolves and potentially adds features, compatibility to evolving OSes and internet standards etc.
3. Provides a platform for immediate application of OpenACS specific patches documentation etc.
4. Provides a strong foundation should aolserver's licence become nonGPL'd at some point.
5. Does not require another domainname registration etc.
6. Keeps the OpenACS community from having to generate and foster another independent developer community (with all of its obligations, responsibilities etc.)
Was the project's development process not suited to foster continued development? What was learned from it?
How would the project be different this time?
In the short term we can get around this by just making tarballs available. In the medium term we might let people upload patches which would be available on an as-is, un-reviewed basis. Ultimately we'd want to have the same sort of process as we have here at OpenACS, where patches are looked at before being accepted. But we haven't got the manpower to make that happen right now.
My main motivation for doing this would be to give us a working example of how subsiting can be used to vastly alter look and feel of said subsites. We don't really have any good real-world example of this out there and I think it's a cool feature.
It doesn't have to be this way, of course, I offer it only as a suggestion.
As far as a distinct project, that's fair. We should get someone to do a quick design for the site, WIHTOUT ANY DOGS!!!
ScottG, you listening? ;)
I will go ahead and register open-nsd.*.
Seriously, there have been previous threads about marketing and AOLserver suggesting that a marketing approach might address any biz-buzz resistance from business types. I'm continuing to gather notes for marketing an AOLserver/OpenNSD configuration.
If anyone intends to stop their marketing creation efforts because of OpenNSD, please post your incomplete work (or forward a copy to me and others), so work can continue using the marketing approach to address AOLserver resistance --your efforts, of course, would/should be recognized in any finished work. These efforts should still be valuable as the material still applies.
I don't think it is using any extra resources. Most of the patches already exist and no one is suggesting patching the 3.x series with new patches.
The official site is extremely user unfriendly, as are all sourceforge sites, so finding the patches is a pain (if they even put them all up there). For example without my UID/GID patch you can't effectivly host multiple sites as different users. The patch is there but it is not at all intuitive to find or even know about.
I will take your word that 4.x will be out soon, but I won't base my plans on it (1. because it is a .0 release and 2. AOL has their own schedule.)
I definitely think we should keep it very, very simple for now - just a little introductory text, a place for people to upload patches, and of course a way to download everything. If 4.0 isn't out by the time we've released OpenACS and dotLRN and caught our collective breaths, then we can talk about doing more. For now, it's mostly just to make resources available, and to get the name OpenNSD on people's minds.
Additionally, a lot of the OpenACS tcl code is in fact very useful in a non-ACS AOLserver environment, but as the code is currently structured, it's a significant pain to actually use it outside of OpenACS.
I would personally like to start refactoring some of
packages/acs-tcl/tcl/*.tcl to work more easily in
non-OpenACS environments, however, I've been refraining from
suggesting it until I actually have the time to do the work.
I've personally successfully used the following ACS Tcl stuff outside of (Open)ACS:
- db_api (and enhanced to support multiple databases)
- Misc. various other utilities
- the Templating system
- possibly the request processor in general, and certainly the part of it which lets you eliminate the .html, .tcl, or whatever file-name extension from your URLs.
- basically anything that doesn't use the OpenACS database schema.
Other OpenACS Tcl code which would be useful outside of OpenACS, but which I've never used, probably includes:
I could be cynical and think that register.com or dotearth.com are selling information on what URLs are being looked up, and that some folks think that's a good way of deciding which expired URLs to buy. I could be even more cynical and think that perhaps the owner, Marcelo Guimaraes, is a reader of this bboard and thinks this is funny? Who knows.
Anyway, I registered open-nsd.org/net/com/info last night, and we'll use those for the new site.
See how easy it is to sell this stuff?
Apache 1 became a messy hairball because they accepted patches based on community votes.
Date: Fri, 20 Sep 2002 11:17:12 -0400
Reply-To: AOLserver Discussion <mailto:AOLSERVER@LISTSERV.AOL.COM>
Sender: AOLserver Discussion <mailto:AOLSERVER@LISTSERV.AOL.COM>
From: Nathan Folkman <mailto:shmooved@MAC.COM>
Subject: Re: [AOLSERVER] Whither the tcl API documentation?
I'm the one who asked Kriston to change aolserver.com to point to SourceForge. Here's why:
- With the exception of the documentation, the rest of the site already pointed to tools on SourceForge, this move just completes the migration.
- While the short term situation with the documentation is admittedly less then ideal, having the documentation administered via SourceForge tools should allow everyone to more easily contribute and update the docs. All of the documentation is still available via a downloadable tar ball which can be installed locally in the mean time.
A quick heads up on some of the work going on here at AOL with regards to AOLserver. You may have noticed an aolserver_v35_bp in CVS. 3.5 is basically the same code base as 3.4.2 but with Tcl ripped out. To compile 3.5, you'll need to check out and compile Tcl 8.4. The whole configure and Makefile setup is identical to what you can expect in 4.0. The 3.5 version was done to help ease the migration to 4.0. Once some more internal testing is complete, we'll tag the code and provide some 3.5 binaries.
Work on AOLserver 4.0 is nearing completion as well. I'll update everyone once we've got some more firm dates. Part of the work going on with 4.0 is to update the Tcl and C documentation.
The contributions by the AOLserver community are valuable and very much appreciated. With the move to SourceForge now complete, there shouldn't be any further changes to the site - now it's up to you! Please feel free to drop me email with any suggestions. Thanks again!
- Nathan Folkman
On Friday, September 20, 2002, at 08:42 AM, Daniel P. Stasinski wrote:
>> I actually liked the aolserver.com site. How was
>> moving to SourceForge a good thing?
> It's been on sourceforge for over a year, just had it's own face. I
> don't think anyone below management level knows why it was changed,
> but considering the flurry of code development going on now, I'm
> remaining happy
> and optimistic.
> Daniel P. Stasinski