Let me throw another thought into this. It is related to the subtle
differences of URLs and URNs, and Bob Wellman's suggestion.
So far we are more or less converging to
protocol://authority/resource_type/resource_id[/cacheable_info]
with the [/cacheable_info] sill up for discussion. Independent on how
you break this down in persistent storage, this is an URL, and one with
specific assumptions about the existence of a resource_id.
However, in the future we would like to interoperate with systems that
aren't virtual worlds as such. In most of those systems, the resources
are named with names that may not even be URLs and even if they are,
they may not have "resource_id" at all. For example,
Cristiano Ronaldo's Facebook handle is http://www.facebook.com/Cristiano
A Collada model has <contributor><author>John.Smith</author></contributor>
(Wondering how Linden Lab is going to deal with this... )
So this would not comply with the form above.
Now, in practice, we have been designing under a fixed-point: the LL
viewer. The LL viewer MUST have UUIDs for every resource that is sent
from the server the client. So, in practice what this means is that
every external resource reference must be assigned a local UUID if it
doesn't come with one, which sounds like a very reasonable thing to do,
and some people here already mentioned it.
Nevertheless, the Collada example breaks the assumption that every named
resource out there has an authority (locator) that we can refer back to.
It may not. We can continue down the road of fixing missing information
with default information. So, for example, if something comes in as
John.Smith, our reference to it could be
http://ourgrid/user/some_uuid/John+Smith -- even though there is no such
account. But that's a slippery slope.
This further supports the argument of having a table that maps between
local UUIDs and external references. And maybe external references may
take more than just one URI form, not just the one we have been talking
about
protocol://authority/resource_type/resource_id[/cacheable_info]
Maybe it can also be a URN.
Nevertheless, the semantics is important, because when an authority
(locator) is present, the receiving grid may want to invoke it.
Melanie wrote:
While you have a case with using a central table, rather than
storing the URL string, and therefore the name, all over the place,
your request schema would not work.
First off, it overcomplicates (IMO) things if you even attempt to
show the current name of an identity. My plan has been to show the
name AT CREATION TIME on a prim, e.g. the displyed creator name of a
prim will not change, even if the underlying identity changes their
name. This removes much complaxity.
Second, your system breaks when a prim is taken to a new grid after
the grid it was created on has vanished from the internet. In that
case, even the initial lookup will fail and you have no data to display.
Therefore, I'd prefer to go with my initial recommendation and
indeed store the URL, including the name, "all over the place". The
client view can always decide to ignore that part and to a table
lookup, or even contact the grid of origin.
It seems that a lot of people here are all for omitting information
that would be helpful for the 90% case in order to make their
particular corner case the norm. 90% of avatars never change names.
Melanie
Bob Wellman wrote:
Can I subvert this discussion back to the original question of Global
Identifiers?
I have been following the discussion on Global Identifiers with interest and I
see strong differences of opinion but merits in both arguments. I agree that
hard coding meaning into the URI seems like an inflexible way forward but I
also see that minimising lookup traffic between grids is desirable too. So I
have been trying to think is there a middle way to compromise on this. I have
the kernel of an idea I would like to put forward for discussion.
Let us say we invent a new table in the local grids database (called UserLookup maybe) and the key to that table is the URI diva proposed (without the user name part) sent intially. Sorry Melanie but please read on before deciding.
Draft table design here:
1) Original URI (key)
2) Original User name
3) Last verification date.
4) Current user name
5) Current URI (optional idea see later)
When the original URI is received (or later referenced) we do a check against this table to see if a lookup is available and proceed as follows:
1) If the lookup is not on file… we immediately send to the other grid for
verification and the user name and we add the lookup to the table with original
user name and current user name the same.
2) If the lookup is on file but out of date … we send to the other grid for
verification (this can be dleayed if need be) and take the latest feedback of
user name and update the lookup with current name.
3) If the lookup is on file but not out of date (most times this will be the
case) … we do nothing.
What constitutes out of date will be configurable in Opensim.ini (out-of-date= x days). In Melanie’s case she will set forever as avatars must not be changeable. For others this can be tailored to their requirements of speed versus flexibility.
The user service already looks up the Prims user UUID to decode the name for local avatars so it just would be changed to do this new lookup at the same time. I think of it this way …The user service currently answer the question “who is this” and would now answer “who in the global community is this”.
This method keeps the database design 3rd normal form rather than storing key decodes all over the database, as to not do this will at some stage cause database chaos just like it did in appearance sometime ago. Can you tell I have a database design background...LOL
If a grid goes bust or is temporary offline we have the last know user info to rely on using just local grid lookup, which covers the very valid argument Melanie made that we need to know identity even if the other grids no longer there.
We can also have the ability to recognise that verification has not been available for a long time so cross grid IM is probably no longer available to this agent. In the case of an avatar moving grid maybe the forwarding address can be sent back and the current URI stored (the fifth column) at the next look up verification. If the forwarding doesnt ecist or is only temporary, it does matter, as we have last known address stored at our grid, which is the best thats possible. Maintaining an IM link of freindship to an agent that moves grids becomes feasible too. So this keeps things flexible for those that think this is the correct approach long term.
In summary this method reduces the need for massive cross grid lookups and still keeps the link between the URI and the user name flexible.
Just my opinion (and I expect to be overruled) but felt I should say this anyway if only to purge my guilt of saying nothing just to avoid a slap down …LOL
Thanks for reading it.. Bob Wellman (PMgrid)
Date: Sun, 29 Aug 2010 19:59:34 -0700
From: [email protected]
To: [email protected]
Subject: Re: [Opensim-dev] Global identifiers
See my previous email about what changed.
We seem to have quite different concepts of what a standards process is.
In my book, a standards process is something that happens *after*
implementations exist, and preferably several competing ones; in the
people in VWRAP's book, it seems to mean "let's design something
together from scratch and on paper".
Let's see how well these two concepts can co-exist. Maybe they can't!
Meadhbh Hamrick wrote:
what's changed?
last year you (diva) seemed to have no interest in merging VWRAP with
Hypergrid. if i remember correctly, hypergrid was going to go off and
do an implementation while VWRAP went another way to try to
incrementally build a standard and some code to implement it.
there's still, of course, room for you at the table. but if i remember
correctly, you seemed to have problems with LLSD/LLIDL (now about to
get renamed DSD) and you had a security model that didn't work with
zha's use cases.
if you're interested in making the case for why VWRAP should adopt the
hypergrid security model and drop DSD, you're welcome to participate
in the VWRAP mailing list.
and i encourage you to actually do so before dropping a spec on the group.
-cheers
-meadhbh
--
meadhbh hamrick * it's pronounced "maeve"
@OhMeadhbh * http://meadhbh.org/ * [email protected]
On Sun, Aug 29, 2010 at 5:29 PM, <[email protected]> wrote:
Melanie is just right, as usual... well, as in 90% of the time :)
The cacheable_data was a brilliant idea, and if you had experience with
OpenSim in the wild, in how volatile these worlds are and how much these
particular remote lookups lag the sims, you would come to appreciate it,
instead of being put off by her dominatrix attitude.
OpenSim core has been contributing center-front in VWRAP: John Hurliman is a
core developer of OpenSim.
Mike Dickson wrote:
That's great to hear. And the first I've heard of it. I'm on the VWRAP
mailing list and yes, John has made some very substantive contributions
to the discussion. I haven't seen anything from OpenSim core during any
part of the discussion to date. I'm a pretty smart guy but not
omnipotent. I've simply interpreted the lack of participation as lack
of interest and past comments would tend to support that (I can dig them
out if you like). And there is no "general feeling" in VWRAP as to your
proposal since its never been presented or discussed there.
I'm not interested in a war, just open dialog and a sincere interest in
interoperability. I'll be glad to read the proposal when its made. In
the meantime I'd appreciate you not attribute negative motives to
anything I've said. I've been simply trying to make technical arguments
against an approach I think is wrong headed and not though out. I've
seen discussion here pretty much get cut off when a core member
"dictates" the solution. Melanie seems to have made up her mind. Fine.
Go build it. Best of luck to you. In the meantime I'll look forward to
the Hypergrid proposal to VWRAP and reserve my comments for that time.
BTW, I've found the VWRAP discussions to be pretty open and devoid of
politics. People will assert politics over almost anything of course
but the dialog has been mostly open and good natured (and quiet lately).
It will be good to have you at the table. Given OpenSim gets a fair bit
of attention it would have been nice if you'd been there all along.
Mike
On Mon, 2010-08-30 at 00:00 +0000, [email protected] wrote:
Mike,
That's an interesting statement to make, considering that John Hurliman
and I are working on writing up the *working* Hypergrid 1.5 as a proposal to
VWRAP, since we have both concluded that the concepts being talked there
lately, without any implementation behind them, are essentially
indistinguishable from the working HG 1.5 that lots of people are already
using.
It seems that you are trying really hard to make this look like a war
between OpenSimulator and VWRAP. I don't think that's the general feeling in
VWRAP, I think it's just you. The proposal to VWRAP will happen. Hopefully,
most people there will be able to assess the technical issues, independent
of the political ones. (emphasis on *hopefully*)
Diva / Crista
Mike Dickson wrote:
Fine, then do what you like. The code's all available. If I don't like
it I can change it. Of course that sort of shoots holes in
interoperability. But then I didn't feel that hyper-grid belonged in
core either for the same reason.
I think you've way over trivialized the whole set of interactions
between agent, asset and simulator services in situations where those
services are defined by different principals. As Meadbh said, this
feels like optimizing to solve a specific problem before you've really
looked at the larger issues. It might be instructive just to simply walk
through some use cases and see where things fall apart. Alot of that
discussion has already taken place on the VWRAP list but OpenSim core
seems to be dead set against involvement in that.
I don't see a way to contribute here beyond the opinion I've already
voiced so I'll drop this.
Mike
On Sun, 2010-08-29 at 22:56 +0000, Melanie wrote:
Sorry, i disagree. The information included is defined by the
REQUIRED data on the recipient, not on what data the sender wants to
provide. the recipient NEEDS a displayable field. It can't be optional.
Melanie
Mike Dickson wrote:
If the decision is to go ahead and do cache-able data then I'd agree,
do
it as attribute NVP's and make them optional. The originating agennt
service is then free to define the semantics of the attributes it
exposes.
Mike
On Sun, 2010-08-29 at 21:42 +0000, Ai Austin wrote:
From: [email protected]
protocol://authority/resource_type/resource_id[/cacheable_data]
+1
consider ensuring that at least the name is provided in a form that
can be resolved fast and locally by including the avatar firstname+lastname
- in whatever form the providing grid wishes to address issues raised by
others - so long as the strings are "legal" in the creator/owner fields.
would it be worth making sure that the "cachable data" is in the form
of keyword=value pairs, and hence put in a "parameter" form after ? rather
than a final /?
protocol://authority/resource_type/resource_id[?key_value_pair[,...]]
with a minimum suggested (or required?)
avatarname=firstname+lastname if the resource_type = user
_______________________________________________
Opensim-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/opensim-dev
_______________________________________________
Opensim-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/opensim-dev
_______________________________________________
Opensim-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/opensim-dev
_______________________________________________
Opensim-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/opensim-dev
_______________________________________________
Opensim-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/opensim-dev
_______________________________________________
Opensim-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/opensim-dev
_______________________________________________
Opensim-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/opensim-dev
------------------------------------------------------------------------
_______________________________________________
Opensim-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/opensim-dev
_______________________________________________
Opensim-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/opensim-dev
_______________________________________________
Opensim-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/opensim-dev
