The email to url resolution thing without a name

[Adewale Oshineye]

I was trying to come up with a session proposal for OpenTech 2009 and
thinking aloud: http://www.jaiku.com/channel/jaikuengine/presence/c6f671f75f744e0b8ea3354b3bc1a912

At which point Brett Slatkin pointed out that the email to url
resolution thing that Blaine was describing to us is an extension of
this: http://brad.livejournal.com/2357444.html with a new describedby
relation. At which point I remembered reading Brad's article last
year.

This kind of identity resolution seems like a good fit for the
SGNodeMapper project: http://code.google.com/p/google-sgnodemapper/
which is the core of the Social Graph API: http://code.google.com/apis/socialgraph/

My overall point is that there seem to be _a lot_ of fragmented
efforts in this area: http://beta.friendfeed.com/kshep/786e9b7a/how-many-open-microblogging-micropublishing
and none of them have really gained any traction. What can or should
be done about this?

[Alexis Richardson]

Ade,

Thanks for bringing this up. I'll assume people are ok with
discussing this here as a general pubsub thing, unless told otherwise.
Although the points you raise are typically discussed in the context
of web content, I think they are in fact almost completely general.

I'll go straight to your last point - leaving the rest to others.

> My overall point is that there seem to be _a lot_ of fragmented
> efforts in this area: http://beta.friendfeed.com/kshep/786e9b7a/how-many-open-microblogging-micropublishing
> and none of them have really gained any traction. What can or should
> be done about this?

IMO the reason for so many fragmented efforts
(micropublishing/microintegration) is as follows:

1. twitter is doing really well and people want some of that
2. there is not consensus on the problem that is being solved
3. so everyone is solving a slightly different problem - dimensions
include federated vs centrally controlled, choice of transport, ..
4. example: numerous ideas about what pubsub is

One theme for a blogpost or talk might be to factor the problem. What
is the social application stack? A couple of months ago Tony and I
met with Blaine to discuss this and - based on notes I took from our
conversation - saw three distinct layers:

A) Social Applications eg content and 'following'

B) Social Networks eg directories and identities

C) Pubsub Networks eg addressing and capabilities

Now, in my view, many of the current efforts take on too much work by
attacking bits of layers A, B and C simultaneously. Moreover when
people start to factor the problem, they disagree about how to do
that. An example of sort-of-but-not-quite related work is here:
http://www.sociallipstick.com/2009/03/04/a-proposal-for-a-conceptual-open-stack/

So, the claim is this. A really good definition of (C) - pubsub -
would significantly lower the cost of attacking A and B for people who
want to do OMB and federated social networks. Pubsub *is* the key.
I'd go further and suggest that a successful model of (C) would be
independent of transport. So for example one might use HTTP when
clients could not keep a socket open, and other protocols such as XMPP
and AMQP in other, connected cases.

Thoughts?

alexis

[Adewale Oshineye]

2009/4/26 Alexis Richardson <alexis.r...@gmail.com>:

I'd like to see where the various implemented and proposed
technologies fit inside that model.
The diagram referred to in this blog post:
http://netmesh.info/jernst/Digital_Identity/proper-open-stack.html has
greater explanatory power because it has examples and it takes into
account the dependencies between the various technologies we have.

I agree with Alexis that a model which talks about passing around a
relatively neutral format (like Atom entries) and left addressing,
identity and transport as mere hooks would be conceptually clean.
However I've had this particular variation of "the 3 blind men and an
elephant" argument at work and ultimately there are too many tribes
who care only about their particular sub-problem. As such conceptual
clarity threatens to undo a lot of existing work and is unlikely to
gain much adoption. The efforts to upgrade to IPv6 or fix fundamental
problems in DNS are exemplars of what happens when conceptual clarity
meets entrenched technology.

By now you've probably all seen Eran's latest draft on the subject of
discovery: http://tools.ietf.org/html/draft-hammer-discovery-03#page-8
This: http://code.google.com/p/jaikuengine/source/browse/trunk/doc/federation_of_planets.txt
is the initial draft describing what JaikuEngine is going to do.

Notice that it cuts across the layers of Alexis's proposed stack
because it has to implement all these layers for itself. Andy Smith,
the main developer on JaikuEngine nowadays, has spoken to the various
other people playing in the area when a lot of them were in San
Francisco for SWFOO. I'm sure someone else has already pointed out the
irony of people working on open protocols in a private invite-only
event which frowns upon live-blogging. Those who weren't there might
find the following blog posts from John Panzer useful:
- http://www.abstractioneer.org/2009/04/personal-web-discovery.html I
strongly recommend this article in particular.
- http://www.abstractioneer.org/2009/04/social-web-foo-standards-for-public.html

I suspect that progress is going to require a self-consciously
Spencerian evolutionary process where we try to ensure that only the
'fittest' technologies and protocols survive. That's going to require
the various different tribes to start having their discussions in the
same place (at the moment each tribe has its discussion space and
largely ignores what everybody else is doing). Only after we start
using the same discussion space can we begin to talk about a common
terminology, goals or tools.

What we need is a public mailing list or some other communication
mechanism where any and all people who interested in _implementing_
these ideas in an interoperable fashion can swap code; test
compatibility and build a network of systems that actually provide
value to end-users. This mechanism would have to explicitly eschew
standards-setting in favour of inclusive experimentation. Otherwise it
will be swamped by bikeshed discussion and vested interests who want
to make sure their tribe's particular sub-problem is resolved and the
'wrong' standards don't win. For instance I'm mostly interested in
getting an eco-system of federated microblogging systems which are
open to radical new ideas because innovation and interoperability can
now co-exist. However there are other people trying to solve related
problems. For instance the Simple Update Protocol guys.

The SoapBuilders mailing list is a good historical example of what I
have in mind. What I don't know is how we make progress in an open and
decentralized fashion whilst also making sure that we don't have
various tribes off building their own incompatible social software
stack because they don't know what everybody else is doing.

I'd appreciate any and all feedback on any of the above.

[Alexis Richardson]

Everyone - please scream if you would rather see this list kept to
Pubsub 'planning'. I can create a group pub-sub-discuss if you'd
rather split off the tech chat.

Ade,

Thanks for this - more below.

On Tue, Apr 28, 2009 at 11:51 PM, Adewale Oshineye <ade...@gmail.com> wrote:
>
>> <snip>

>
> I'd like to see where the various implemented and proposed
> technologies fit inside that model.
> The diagram referred to in this blog post:
> http://netmesh.info/jernst/Digital_Identity/proper-open-stack.html has
> greater explanatory power because it has examples and it takes into
> account the dependencies between the various technologies we have.

That stack is finer grained and potentially conflates several kinds of
addressing in the bottom layer. I think the four upper layers make
some sense and could sit above pubsub, which is integrally commingled
with addressing. But ... yes.

> I agree with Alexis that a model which talks about passing around a
> relatively neutral format (like Atom entries) and left addressing,
> identity and transport as mere hooks would be conceptually clean.
> However I've had this particular variation of "the 3 blind men and an
> elephant" argument at work and ultimately there are too many tribes
> who care only about their particular sub-problem. As such conceptual
> clarity threatens to undo a lot of existing work and is unlikely to
> gain much adoption. The efforts to upgrade to IPv6 or fix fundamental
> problems in DNS are exemplars of what happens when conceptual clarity
> meets entrenched technology.

One thing I admire about the work at pubsubhubbub is that it does not
run the one against the other, yet it is reasonably clear at the same
time.. I agree that IPv6 is a great example of why fixing via an
overlay is easier than changing the 'underlay' ... Personally I
believe that the various approaches to pubsub have a role in that
overlay. Of which more on request ;-)

> By now you've probably all seen Eran's latest draft on the subject of
> discovery: http://tools.ietf.org/html/draft-hammer-discovery-03#page-8
> This: http://code.google.com/p/jaikuengine/source/browse/trunk/doc/federation_of_planets.txt
> is the initial draft describing what JaikuEngine is going to do.
>
> Notice that it cuts across the layers of Alexis's proposed stack
> because it has to implement all these layers for itself.

Another reason it appears to do this is because it recycles the same
addressing model (afaict at first glance) between the application,
network and pubsub layers.

> Andy Smith,
> the main developer on JaikuEngine nowadays, has spoken to the various
> other people playing in the area when a lot of them were in San
> Francisco for SWFOO. I'm sure someone else has already pointed out the
> irony of people working on open protocols in a private invite-only
> event which frowns upon live-blogging.

Tell me about it ;-)

> Those who weren't there might
> find the following blog posts from John Panzer useful:
> - http://www.abstractioneer.org/2009/04/personal-web-discovery.html I
> strongly recommend this article in particular.
> - http://www.abstractioneer.org/2009/04/social-web-foo-standards-for-public.html

Nice write ups.

Reusing URLspace for names would enable a persistent public social
graph suitable for persistent public media (eg facebook) but federated
(ie not locked in to facebook).

I am hopeful Blaine will now chip on how to make this more memorably
addressable for humans (or pace your Jaiku galactic federation,
humanoids). It seems overlap with Eran's IETF draft can be
anticipated - but I have only skimmed the docs.

> I suspect that progress is going to require a self-consciously
> Spencerian evolutionary process where we try to ensure that only the
> 'fittest' technologies and protocols survive. That's going to require
> the various different tribes to start having their discussions in the
> same place (at the moment each tribe has its discussion space and
> largely ignores what everybody else is doing). Only after we start
> using the same discussion space can we begin to talk about a common
> terminology, goals or tools.

I and I think others on this list believe that the study and
implementation of pubsub systems is a good slice through this problem.
JaikuEngine is an example of 'a pubsub system' but I think the story
needs to be a bit more general (but not very much more so) to see the
common patterns. They can then be applied to other problems including
federation of social apps.

> What we need is a public mailing list or some other communication
> mechanism where any and all people who interested in _implementing_
> these ideas in an interoperable fashion can swap code; test
> compatibility and build a network of systems that actually provide
> value to end-users.

+1 and discuss them

> This mechanism would have to explicitly eschew
> standards-setting in favour of inclusive experimentation.

+1 we have enough standards. It's the patterns and metapatterns that matter.

> Otherwise it
> will be swamped by bikeshed discussion and vested interests who want
> to make sure their tribe's particular sub-problem is resolved and the
> 'wrong' standards don't win. For instance I'm mostly interested in
> getting an eco-system of federated microblogging systems which are
> open to radical new ideas because innovation and interoperability can
> now co-exist. However there are other people trying to solve related
> problems. For instance the Simple Update Protocol guys.

Yes.

> The SoapBuilders mailing list is a good historical example of what I
> have in mind. What I don't know is how we make progress in an open and
> decentralized fashion whilst also making sure that we don't have
> various tribes off building their own incompatible social software
> stack because they don't know what everybody else is doing.
>
> I'd appreciate any and all feedback on any of the above.

It sounds great to me. I hope others comment.

We started Pubsub to talk about this over beers. This list could be a
place to discuss online, or we could make a pub-sub-discuss for people
to do that. Would that be useful?

alexis





> >
>

[Alexander Jones]

Scream!