Forum OpenACS Development: General Comments and subsites

Collapse
Posted by Pascal Scheffers on
[This is a transcript of three emails by me and Michael Steigman]


Hi Pascal,

I'm trying to get the general comments package working with some domain based subsites and, seeing as how you were the porter, I thought I'd check with you on a few questions I had (though I'm using Oracle). I guess my first question would be is it possible to set up the above without doing any hacking? I looked at the data model and some of the procs and it doesn't look like the package is subsite aware yet I've heard from some that GC works on subsites if you mount GC under each subsite. I don't see how that would work though.

If there is no subsite support, was there a reason for that decision (other than lack of time)? I ask because it doesn't *seem* like it would be too tough to add subsite support (though I haven't looked at static pages). Also, as I was looking through the procs file, I noticed some deprecated procs that look like they contain the remnants of some subsite support. Is that what it looks like?

Thanks.

Michael


Hi Michael,

On Thu, 2002-04-11 at 17:48, Michael Steigman wrote:

(The delay was because I was on vacation)
>I'm trying to get the general comments package working with some domain
>based subsites and, seeing as how you were the porter, I thought I'd
>check with you on a few questions I had (though I'm using Oracle). I

I'm afraid (check the date stamps) that most work I did is nearly one year old. The porting I did works for PG but is not 100% (the content-repository wasn't done back then), and the Oracle code was never updated. Sorry. I changed employers and that left me without time to do more on 4.x.

IIRC, subsite stuff does indeed _not_ work for GC, feel free to take over the work if you want to.

Regards,
- Pascal.


Thanks for getting back to me, Pascal. I figured you might be on vacation.

I noticed that the code still uses the acs-message interface (whereas Jon Griffin mentioned in a bboard posting that GC should call the CR API directly). Your comment explains why that is. After looking over the code, I made a few changes and created a bug and patch at:

https://openacs.org/sdm/one-baf.tcl?baf_id=1478

If you have time to give it a test, I'd love some feedback. As far as the rest of the unfinished work is concerned, if you can recall any more particulars about what was generally agreed upon or desired for the data model and pages, I can probably work on that stuff (time permitting). Thanks.

Michael

Collapse
Posted by Pascal Scheffers on
Like I said, my last work on GC is a bit old. The patch seems okay, but I don't have an anywhere near current OpenACS 4 install anymore (last update was in september IIRC) - so I can't really test it.

IIRC, the oracle stuff was done by the query-extractor with a cursory glance/test for correctness (just a test run, which did cover most of the normal code, but few of the special cases), there may even be FIXMEs for Oracle in there (probably not).

The most important things were the using the CR, which was never finished: URL attachments work, others don't (I may have started on that). There are one or two places (small) where the Oracle code is not in sync with the PG code. That is about all I can remember, and about all there was to it as well.

Collapse
Posted by Don Baccus on
I'll take a look at the patch.  Also, I did some clean-up work on general comments late last year.

Good to hear from you, Pascal, we miss you :)

Collapse
Posted by Pascal Scheffers on
> Good to hear from you, Pascal, we miss you :)
Thanks :)

I'm still working with ACS 3.2.5. I've been working (alone) on a respondent tracking system for Generation R, a population study of the Erasmus University Rotterdam. We've begun to keep a tag on 10.000 children and their parents, starting at 12th week pregnancy following them at least until they're 18. We're at about 500 right now (as far as I can tell). The system is working.

It prints (latex mailmerge and numbered postscript questionaires), it tracks, it warns, it keeps stock, it anonymizes data, it keeps track of consent status, its configurable, can be used by multiple studies at the same time. Next up is datamanagement, to preprocess the raw study data sources/import filters (things like TONS of questionaires with 20 pages each, medical exam data, ultrasound reports, that sort of stuff) into something a bit more palatable by SPSS and SAS.

I've got permission to release the code, but I need to find a weekend to translate the interface to english, it's in Dutch right now :(

I've missed ACS a bit too - 've been a bit busy with a lot of other things, like cryptorights.org, too much travel (the get-to-work type), diving, ...

Collapse
Posted by Roger Williams on
I am unclear about what is the user/application requirement for a sub-sites enabled GC package. I guess I thought users cannot tell where (or how many) GC packages are mounted, nor need they care. In our site, the comments are like attachments to the underlying object they are referring to. No?

Please broaden my view. TIA.

Regards..

Collapse
Posted by Don Baccus on
I don't know if I'll have time to play with this soon or not ... comments displayed by client packages (say, file storage) shouldn't need to know where the general comments package is mounted, which is the point Roger's getting at.

However ... a subsite-scoped search of the CR will need to know this information.

I think this is an example of a more general problem with CR content but haven't really looked at this ...

Collapse
Posted by Michael Steigman on
I guess I thought users cannot tell where (or how many) GC packages are mounted, nor need they care. In our site, the comments are like attachments to the underlying object they are referring to. No?

I think you are right, though the "view details" link in the comments display of the current implementation *does* need to know where GC is mounted. The main reason I started hacking GC, though, had nothing to do with displaying comments - it was because I could not get the "Add a comment" link to work correctly with subsites (both the garden variety and domain based). While I was at it, I thought it'd be nice to have the ability to create discrete GC instances. I am planning to set up a single OACS installation with domain based subsites and I'd like to be able to delegate the management of comments on different sites to different people. I guess that would be the requirement, in answer to Roger's question. Does that make sense?

So the impetus behind the patch was to correct the general_comments_package_url proc, which was just wrong. The package_id/subsite stuff just seemed like low-hanging fruit. It was basically modeled on what I saw in the News package.

Collapse
Posted by Don Baccus on
Yes, this does make sense.  Being able to delegate administration in this way is certainly a requirement.

I'll try to find time to look over your patch in the next few days, unfortunately I'm swamped with pay-the-rent work at the moment ...