Actually, we need both. It's perfectly valid to index and address users by guid 
in the known internal scope - but conversely, invalid to refer by guid in a 
cross-scope case. Also, the guid contains no 'resource location' data, which 
urls are meant to provide.

 

So, A local user would have a guid ID at any rate.

 

The hashing algorithm seems fair enough to me, IE, hash the url with a 
scope-wide secret salt, and use that only for internal indexing, ie, fetch the 
user profile from the url, but assign it the salted hash guid as ID internally. 
Keep in in-memory on-demand and never write it to db. Of course, this means 
you'll have to make sure all references to it is by url, not id.


Incidentally, I'm thinking we could do the same for assets - let regions 
substitute urls embedded in assets with hashed salted guids, allowing us to 
reference web resources (jpegs/pngs) in other assets - and if needed for a 
particular protocol, autoconvert them to, say, j2000. The general idea is to 
treat what is now the grid scope as the region scope, essentially converting 
the region to a cache/conversion mechanism for the wider web.

 

And yes, I'm not saying there will not be problems. There will be tons of them. 
But again, think of the possibilities. One could start by solving one case, 
then the next, and so on. :-D


Best regards,
Stefan Andersson
Tribal Media AB



 
> Date: Mon, 30 Mar 2009 08:08:30 -0400
> From: [email protected]
> To: [email protected]
> Subject: Re: [Opensim-dev] RFC: Ways of creating profiles for creators who 
> will never log in
> 
> Melanie wrote:
> > I am absolutely sure that I would not went HG to ever, EVER, insert 
> > user records or profile records into my database. I strongly favor 
> > the "URL" approach, which is neutral as far as my database goes.
> 
> Ok, so then hash those urls for UUIDs to present to the viewer? We
> still need a UUID in the object definitions for the viewer so that
> people can see the creator.
> 
> -Sean
> 
> -- 
> Sean Dague / Neas Bade
> [email protected]
> http://dague.net
> 
> 
_______________________________________________
Opensim-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/opensim-dev

Reply via email to