Expansion of FOAF support

Feb 24, 2004 08:20

What data would you like to see included in the FOAF spec next? What should or could LiveJournal support, what additions could be made, what else woudl fit into our data set ( Read more... )

Leave a comment

Comments 27

(The comment has been removed)

crschmidt February 24 2004, 06:50:35 UTC
For users who publically list contact information, this is already done:

47729022
cr5chmidt
crschmidt319@msn.com
crschmidt

We obviously don't want to show information that the user has chosen not to list publically. Is there anything wrong with the information that's already there?

Reply

(The comment has been removed)

crschmidt February 24 2004, 07:03:40 UTC
No problem - the contact info kind of piles up on top of the interests to the eye: I coded it, and I missed it myself when I looked :)

Reply


breyten February 24 2004, 08:04:56 UTC
Well, I found that foaf:depiction, nor foaf:img is supported (I was a bit surprised), so I would like to see that (not only for the main person, but also for the foaf:knows).

Reply

crschmidt February 24 2004, 08:11:17 UTC
Is this an accurate foaf:depiction? How can I tell what is? How do I know whether an icon is an accurate foaf:depiction field or not?

Default userpictures are not a valid example of foaf:depiction, sadly, and as such, there isn't really a good, easy way of doing this.

Additionally, VERY minimal data is loaded for foaf:knows. If you want to find out about that person, you need to go to THEIR foaf file, because loading 750-1000 userpic information sets, even with the caching LJ uses, is not quick enough, and would be a major DB hit. FOAF files will likely never load extra data about foaf:knows, as it simply can't be done with the current setup.

Reply

theorb February 24 2004, 08:39:24 UTC
foaf:icon would seem to be the proper thing to use here. Your userpic is used as a visual indicator of you, even though you may or may not appear in it. Similarly, the logo for, say, a pentium 3 is not a picture of a P3.

(The major difference is that most logos attempt to be /identifing/ marks -- that is, given a logo, it should represent only one Agent. That's not true of LJ userpics; it isn't at all uncommon for more then one person to have the same, or at least very similar, userpics. However, "attempt" is important here.)

Reply

breyten February 24 2004, 09:11:09 UTC

You can't tell. So the consensus is that depictions need to be real? Maybe a foaf:avatar would be more appropriate for Livejournal.

Default userpictures are not a valid example of foaf:depiction, sadly, and as such, there isn't really a good, easy way of doing this.

I see. Thanks for explaining :). You could make it so that attaching the keyword 'foaf' to a userpic would result in that pic being used as such, but that's still a bit ugly :)

you want to find out about that person, you need to go to THEIR foaf file

You're absolutely right. That was a bit of a braino :)

Reply


theorb February 24 2004, 08:26:02 UTC
First off, I'd like to note that I am /not/ writing anything that uses FOAF, nor am I likely to.

However, if I were, some things I'd like to see that don't seem to be there:
1) Support for geographic information other then lat-long. For example, userinfo shows my location as München, Bayern, Germany. While this is hardly exact (it could have been Munich, Baviaria, Germany, and refered to the same place), it's a lot better then the way it's currently rendered in the FOAF data -- which is to say, not at all. I don't see anything for this in the FOAF spec, but then again I don't see the lat-long stuff used in at least one FOAF example I saw there either.

2) It may be useful to give a foaf:logo of the person's default image. This would seem to match usage on the userinfo page, and may be useful to clients that give maps (weather based on physical location or network diagrams).

Reply

crschmidt February 24 2004, 10:22:59 UTC
The location stuff for FOAF thus far is all either
1. Airport based. (basedNearAirport)
2. Geo Based (Lat/Lon)

Neither of these is really ideal, because they aren't easily matchable to what we have (Not many geo tools out there do Lat/Lon on anything less than a street address, and nearAirport is difficult in all cases).

This sounds like something that would need to be a modification of the FOAF spec in order for us to put any usable data in it - you may want to bring this up to them. At this point, I'm mostly using existing FOAF fields and mapping them to what we've got, simply for ease of implementation.

The FOAF logo field is dreadfully underspecified - until I see it included somewhere, I'm not sure how to use it. Do you have an example? http://xmlns.com/foaf/0.1/#term_logo is the xmlns definition, which is rather useless :)

Reply

surrealpenguin February 24 2004, 14:57:56 UTC
If you want to write a little more code (who doesn't!), then you can maybe translate city names to lat/long with a good database, perhaps like the free version of this thing.

Reply

crschmidt February 24 2004, 15:07:44 UTC
Problem is that we can't really do much processing on the fly. Anything we do would have to be very minimal (MD5ing a page would be alright, doing another database query might not be, etc. etc.)

Because we want these profiles to be easily spiderable, we want to make sure they're easy on us. THis means not doing any advanced mathematics or computations, in general, so we wouldn't want to add something that might be heavy weight.

Do you know of any free, already coded, lightweight code that does this?

Reply


freso February 24 2004, 12:07:47 UTC

  1. I think it was Leigh Dodds who mentioned this in #foaf, and I thought I'd bring it up here - how about adding something like this:

    freso



    to the LJ FOAF files?
  2. How about being able to make an external FOAF-file, making an rdf:seeAlso reference and keeping the rest of the LJ FOAF-file intact? So I could use the LJ FOAF to point to the FOAF in my personal webspace, while having the FOAF in my personal space pointing at the LJ FOAF. That way, I wouldn't have to manually keep track of changes to friends list, interests, and what not, and updating them into the external FOAF... ?

Reply


mcfnord February 24 2004, 19:49:55 UTC
I would very much welcome addition of the self-described Location information. It pains me if such data would violate some central principle of FOAF. Currently I screenscrape the who-i-read data, and the Location, for the meme project at http://ljmindmap.com. I am on hold with the project because last night I tested the automated scraper, and seemed to get blocked temporarily from livejournal.com. I will try again, but I am hunching it was a message and the message is: "Don't do that to us." Yet is there any other interface available yet?

I wrote to brad and evan about getting the who-i-read and Location data on cd-rom, but did not get a reply. Currently I have no way to keep up with the request pace when my meme is live. Your FOAF work is THIS close, but Location is used as a hovertip in the renderings, and I don't want to give it up.

I could give you ten bucks.

Is that a valid reason to evolve a spec? ;)

Reply

crschmidt February 24 2004, 21:34:59 UTC
Heh ( ... )

Reply

mcfnord February 24 2004, 21:54:18 UTC
Thanks for the reply.

I emailed Brad (and Evan) two days ago, but have not heard back. It is not so easy to find some of his time.

I am going to use your FOAF, and scrape the Locations I need in a second step, as I don't need Location for all users. Your implementation gives me datasets for popular users (like hepkitten, gynocide, many others) that scraping can't achieve. If I had Location, I'd quit scraping entirely, and do a happy dance.

Reply

theorb February 25 2004, 02:02:14 UTC
I think that giving a us, and a 17601. Country would be defined to be one of the ISO-3661 country codes. Postcode would be defined to be whatever method of postal code is common in the named country, or (for many countries?) a prefix thereof, so long as that prefix is not unto itself a valid post-code of the same precision ( ... )

Reply


Leave a comment

Up