OpenID Support

classic Classic list List threaded Threaded
13 messages Options
Reply | Threaded
Open this post in threaded view
|

OpenID Support

Evan Prodromou-3
Brion said:

    /It doesn't solve our specified requirement (single Wikimedia-wide
    namespace so a
    single username works on all 600+ Wikimedia wikis transparently, no
    muss no fuss).
    /

Really? I'd think that an Interwiki-namespaced username would solve that
problem nicely, without requiring a big name-clash-fixing step when the
technology rolls out. I think that having a single username across all
wikis would be nice, if you didn't have to do resolve dupes across the
system, but if you do, it's really kind of a hassle.

I'd think that, since single-signon between wikis is a feature useful
only for a minority of registered Wikimedia users (how many actually log
into more than one Wikimedia wiki, ever? 10-20%, maybe? how many log
into more than, say, 5 wikis? Or log into more than one wiki on a
regular basis? 2%? 0.2%?), it'd be a good political idea to minimize the
namespace-sorting hassle. That nameclash-fixing step is gonna suck, and
it's not even helpful most users. It seems to me that a minor effort on
the part of "interwikiists" -- just using a project+language namespace
-- would be painless for them and unnoticeable for everyone else.

I think that if we use Interwiki prefixes on the UI side, OpenID becomes
that much easier. A user could login to French Wikibooks as
/wp:en:User:EvanProdromou/ or whatever, and the UI translates that into
the right OpenID URL (http://en.wikipedia.org/wiki/User:EvanProdromou)
and goes through the OpenID two-step to get credentials checked. Under
the covers, the OpenID stuff would get worked out right, but we could
use simpler Interwiki strings for the UI.

In fact, a drop-down box for project, and another for language, could
also significantly reduce the complexity for users. For example, I could
choose "Wikipedia" out of "Wikipedia, Wikibooks, Wiktionary, ..." and
"English" out of "English, Fran├žais, Esperanto, ..." and then my
username on that project.

Lastly, I think having a couple of professional developers eagerly
awaiting a chance to get this implemented and who know OpenID and
authentication issues inside and out is a really good thing. We
shouldn't rollout every technology that people volunteer to throw into
the software, but this seems like a pretty good match between Mediawiki
(and Wikimedia) needs and what's being offered.

~Evan

P.S. Sorry about the broken threading -- I had some mail server problems
and missed my daily dose of wikitech-l.

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: OpenID Support

Brion Vibber
Evan Prodromou wrote:
> Really? I'd think that an Interwiki-namespaced username would solve that
> problem nicely,

No, it solves a different problem.

> without requiring a big name-clash-fixing step when the

That step is inherent in the requirement of a single namespace.

-- brion vibber (brion @ pobox.com)


_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l

signature.asc (257 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: OpenID Support

Brion Vibber
Brion Vibber wrote:
> Evan Prodromou wrote:
>> Really? I'd think that an Interwiki-namespaced username would solve that
>> problem nicely,
>
> No, it solves a different problem.
>
>> without requiring a big name-clash-fixing step when the
>
> That step is inherent in the requirement of a single namespace.

Sorry to be short about this, but we've had enough fights over this issue in the
past, and as far as I'm concerned the decisions have been made. I'll get flak no
matter whether I implement the previously-agreed conversion specification or
something new, so it's lose-lose for me personally. ;)


However to summarize, my view on why a single namespace is Good includes a few
points:


* Encourages community cohesiveness across diverse projects:

Many people are active in multiple projects. While they may be numerically a
minority, these are among the most active contributors, and thus key to the
projects' success.

One of the things making it hard now to pop over to another projects is that you
have to register another account; if you instead have a "foreign" account saying
you're from another web site, this makes you stand out even more as not belonging.

We want the Wikimedia projects to be able to share people and content easily and
freely. Being an equal user on every project helps break down those borders,
where having a freakish foreign username would build them up.


* Related to that, we have projects specifically for sharing:

Commons and Meta are important. They should be as closely integrated as possible
to make media content sharing and community-wide issue discussion as easy to
dive into as possible. Being given a foreign name when you go there makes it a
foreign place instead of just a special part of the wiki.


* Avoids the 'tld problem':

Even though we all know top-level domains in the DNS system are supposed to be
disambiguators, everybody ends up trying to register their name in every
possible domain -- if you don't, the others mostly go to squatters, SEO scum,
and phishers who have no legitimate claim to the name.

The disambiguators are unsatisfying, and aren't necessarily enough to pass
casual inspection -- especially when you legitimately have a hundred accounts
named 'Angela' or 'Brion Vibber' which _mostly_ are the same person already.

It makes more sense in my view to recognize that the disambiguators don't work
that well when they're needed and make more trouble when they're not, and just
jump straight to a single namespace.


When we add OpenID/whatever in the future later, that'll allow identification
from 'foreign' sites in a limited way which is useful. (Eg a post on the
LiveJournal article claiming to be from Brad Fitzpatrick could be verified as
really having from from Brad's LJ account, and a post on LJ about Wikipedia
claiming to be from Jimbo Wales could be verified as really coming from Jimmy's
Wikimedia account.) But that works because it's *all about* the boundary, and
drawing attention to the foreign boundary is what you want in that case. That's
not, I think, what we want inside Wikimedia.

-- brion vibber (brion @ pobox.com)


_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l

signature.asc (257 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: OpenID Support

Rob Lanphier
As one of the past advocates for OpenID as a possible solution to the
"single signon problem", I wholeheartedly agree with what Brion wrote
below.  OpenID solves an interesting problem, just not the problem that
the most active contributors to the site seem to be interested in
solving.

The spec below tries to address the problem head on:
http://meta.wikimedia.org/wiki/Single_login_specifications

Rob

On Tue, 2006-02-07 at 19:34 -0800, Brion Vibber wrote:

> Brion Vibber wrote:
> > Evan Prodromou wrote:
> >> Really? I'd think that an Interwiki-namespaced username would solve that
> >> problem nicely,
> >
> > No, it solves a different problem.
> >
> >> without requiring a big name-clash-fixing step when the
> >
> > That step is inherent in the requirement of a single namespace.
>
> Sorry to be short about this, but we've had enough fights over this issue in the
> past, and as far as I'm concerned the decisions have been made. I'll get flak no
> matter whether I implement the previously-agreed conversion specification or
> something new, so it's lose-lose for me personally. ;)
>
>
> However to summarize, my view on why a single namespace is Good includes a few
> points:
>
>
> * Encourages community cohesiveness across diverse projects:
>
> Many people are active in multiple projects. While they may be numerically a
> minority, these are among the most active contributors, and thus key to the
> projects' success.
>
> One of the things making it hard now to pop over to another projects is that you
> have to register another account; if you instead have a "foreign" account saying
> you're from another web site, this makes you stand out even more as not belonging.
>
> We want the Wikimedia projects to be able to share people and content easily and
> freely. Being an equal user on every project helps break down those borders,
> where having a freakish foreign username would build them up.
>
>
> * Related to that, we have projects specifically for sharing:
>
> Commons and Meta are important. They should be as closely integrated as possible
> to make media content sharing and community-wide issue discussion as easy to
> dive into as possible. Being given a foreign name when you go there makes it a
> foreign place instead of just a special part of the wiki.
>
>
> * Avoids the 'tld problem':
>
> Even though we all know top-level domains in the DNS system are supposed to be
> disambiguators, everybody ends up trying to register their name in every
> possible domain -- if you don't, the others mostly go to squatters, SEO scum,
> and phishers who have no legitimate claim to the name.
>
> The disambiguators are unsatisfying, and aren't necessarily enough to pass
> casual inspection -- especially when you legitimately have a hundred accounts
> named 'Angela' or 'Brion Vibber' which _mostly_ are the same person already.
>
> It makes more sense in my view to recognize that the disambiguators don't work
> that well when they're needed and make more trouble when they're not, and just
> jump straight to a single namespace.
>
>
> When we add OpenID/whatever in the future later, that'll allow identification
> from 'foreign' sites in a limited way which is useful. (Eg a post on the
> LiveJournal article claiming to be from Brad Fitzpatrick could be verified as
> really having from from Brad's LJ account, and a post on LJ about Wikipedia
> claiming to be from Jimbo Wales could be verified as really coming from Jimmy's
> Wikimedia account.) But that works because it's *all about* the boundary, and
> drawing attention to the foreign boundary is what you want in that case. That's
> not, I think, what we want inside Wikimedia.
>
> -- brion vibber (brion @ pobox.com)
>
> _______________________________________________
> Wikitech-l mailing list
> [hidden email]
> http://mail.wikipedia.org/mailman/listinfo/wikitech-l

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: OpenID Support

Gerard Meijssen-3
In reply to this post by Evan Prodromou-3
Evan Prodromou wrote:

> Brion said:
>
>    /It doesn't solve our specified requirement (single Wikimedia-wide
>    namespace so a
>    single username works on all 600+ Wikimedia wikis transparently, no
>    muss no fuss).
>    /
>
> Really? I'd think that an Interwiki-namespaced username would solve
> that problem nicely, without requiring a big name-clash-fixing step
> when the technology rolls out. I think that having a single username
> across all wikis would be nice, if you didn't have to do resolve dupes
> across the system, but if you do, it's really kind of a hassle.
>
> I'd think that, since single-signon between wikis is a feature useful
> only for a minority of registered Wikimedia users (how many actually
> log into more than one Wikimedia wiki, ever? 10-20%, maybe? how many
> log into more than, say, 5 wikis? Or log into more than one wiki on a
> regular basis? 2%? 0.2%?), it'd be a good political idea to minimize
> the namespace-sorting hassle. That nameclash-fixing step is gonna
> suck, and it's not even helpful most users. It seems to me that a
> minor effort on the part of "interwikiists" -- just using a
> project+language namespace -- would be painless for them and
> unnoticeable for everyone else.
Given that Commons should be the only resource for pictures, everyone
needs currently at least TWO users. Bringing this as a "political" idea
is horrible certainly when you do not consider that several projects do
not allow upload to the local project any more. It is horrible because
it shows single project/language focus. When you consider the current
interwiki practice, it is horribly broken; I have only on the English
Wiktionary 15306 edits doing interwiki stuff for Februari 2006. I am
certain that I have some 50.000 edits on all Wiktionaries only in
Februari. This idea that projects and its language versions are not
related is wrong.

When we get Wikidata databases that are properly localised, we will be
able to have information about one topic in one database available for
ALL projects, this will be another example where interaction between
projects will become relevant.
>
> I think that if we use Interwiki prefixes on the UI side, OpenID
> becomes that much easier. A user could login to French Wikibooks as
> /wp:en:User:EvanProdromou/ or whatever, and the UI translates that
> into the right OpenID URL
> (http://en.wikipedia.org/wiki/User:EvanProdromou) and goes through the
> OpenID two-step to get credentials checked. Under the covers, the
> OpenID stuff would get worked out right, but we could use simpler
> Interwiki strings for the UI.
OpenID is a great idea, it is included in the YADIS specification, a
framework where it interacts with more protocols than just OpenID. The
problem is that OpenID on its own will not provide the needs that the
Wikimedia Foundation will have. When we get Wikiversity running for
instance, we will have teachers that need to have access to the process
information of a students training. This is private information and it
requires stronger authentication than the authentication offered by OpenID.

>
> In fact, a drop-down box for project, and another for language, could
> also significantly reduce the complexity for users. For example, I
> could choose "Wikipedia" out of "Wikipedia, Wikibooks, Wiktionary,
> ..." and "English" out of "English, Fran├žais, Esperanto, ..." and then
> my username on that project.
>
> Lastly, I think having a couple of professional developers eagerly
> awaiting a chance to get this implemented and who know OpenID and
> authentication issues inside and out is a really good thing. We
> shouldn't rollout every technology that people volunteer to throw into
> the software, but this seems like a pretty good match between
> Mediawiki (and Wikimedia) needs and what's being offered.
>
> ~Evan
It is great when developers want a chance to implement something good.
But do not approach it with this "all I have is a hammer so all my
problems are nails" idea. OpenID is great, it fits into YADIS,
technically it is possible to make a WMF authentication server BUT this
is not only a technical issue it is indeed an organisational issue.
Think of it, WillieOnWheels has a profile with the WMF, he gets banned
and still authenticates using an OpenID/YADIS profile ...

It is a technical issue if you want to implement YADIS for MediaWiki. It
is however a decision for the board if the WMF would and should turn it on.

Thanks,
     GerardM
_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: OpenID Support

Josh Hoyt-2
In reply to this post by Rob Lanphier
On 2/7/06, Rob Lanphier <[hidden email]> wrote:
> The spec below tries to address the problem head on:
> http://meta.wikimedia.org/wiki/Single_login_specifications

Thanks for that link. I'm still digesting this whole issue, but it
seems to me that most of the discussion about single-sign-on is about
the namespace, and the choice of sign-on protocol seems like a
separate issue. I don't have an opinion on the namespace issues, and
if I did, it wouldn't matter, because I have not been involved in that
discussion. I do have knowledge about protocols, and what I am
suggesting is that OpenID is a technically appropriate protocol to use
regardless of intended user experience and regardless of the political
and social issues.

I will be educating myself on the discussions that have already taken
place, and commenting further once I have a better understanding.

Josh
_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: OpenID Support

Ryan Lane
In reply to this post by Brion Vibber
Brion Vibber <brion@...> writes:

>
> Brion Vibber wrote:
> Sorry to be short about this, but we've had enough fights over this issue in
> the past, and as far as I'm concerned the decisions have been made. I'll get
> flak no matter whether I implement the previously-agreed conversion
> specification or something new, so it's lose-lose for me personally. ;)
>
> However to summarize, my view on why a single namespace is Good includes a few
> points:

What do you plan on using for the single-sign-on transition (from a
technological standpoint)? Is this posted somewhere? I wasn't aware until
recently that this would be happening soon. This discussion sparked my curiosity.

Ryan Lane

_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: OpenID Support

Nikola Smolenski
In reply to this post by Evan Prodromou-3
Evan Prodromou wrote:
> I'd think that, since single-signon between wikis is a feature useful
> only for a minority of registered Wikimedia users (how many actually log
> into more than one Wikimedia wiki, ever? 10-20%, maybe? how many log
> into more than, say, 5 wikis? Or log into more than one wiki on a
> regular basis? 2%? 0.2%?), it'd be a good political idea to minimize the

It is my experience that people who learned English as a second language
often edit on regular basis both their language and English Wikipedia. I
have no idea how much of them are there, but certainly more than 2%.

Also, it is likely that whatever the percentage is, it will increase
after single sign-on is introduced. We want to encourage people to
upload to Wikimedia Commons...

> I think that if we use Interwiki prefixes on the UI side, OpenID becomes
> that much easier. A user could login to French Wikibooks as
> /wp:en:User:EvanProdromou/ or whatever, and the UI translates that into
> the right OpenID URL (http://en.wikipedia.org/wiki/User:EvanProdromou)
> and goes through the OpenID two-step to get credentials checked. Under
> the covers, the OpenID stuff would get worked out right, but we could
> use simpler Interwiki strings for the UI.

I could agree with this however. I see no problem on the user side with
prefixes.
_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: OpenID Support

Gerard Meijssen-3
Nikola Smolenski wrote:

> Evan Prodromou wrote:
>> I'd think that, since single-signon between wikis is a feature useful
>> only for a minority of registered Wikimedia users (how many actually
>> log into more than one Wikimedia wiki, ever? 10-20%, maybe? how many
>> log into more than, say, 5 wikis? Or log into more than one wiki on a
>> regular basis? 2%? 0.2%?), it'd be a good political idea to minimize the
>
> It is my experience that people who learned English as a second
> language often edit on regular basis both their language and English
> Wikipedia. I have no idea how much of them are there, but certainly
> more than 2%.
>
> Also, it is likely that whatever the percentage is, it will increase
> after single sign-on is introduced. We want to encourage people to
> upload to Wikimedia Commons...
>
>> I think that if we use Interwiki prefixes on the UI side, OpenID
>> becomes that much easier. A user could login to French Wikibooks as
>> /wp:en:User:EvanProdromou/ or whatever, and the UI translates that
>> into the right OpenID URL
>> (http://en.wikipedia.org/wiki/User:EvanProdromou) and goes through
>> the OpenID two-step to get credentials checked. Under the covers, the
>> OpenID stuff would get worked out right, but we could use simpler
>> Interwiki strings for the UI.
>
> I could agree with this however. I see no problem on the user side
> with prefixes.
Hoi,
Authentication is a hot topic, one of the more recent interesting bits
of information can be found here:
http://dystopics.dump.be/2006/02/04/the-mysteries-of-x-google-token-and-why-it-matters/
It is not YADIS or OpenID but I think that it may point to thinks that
are in the future, but it is certainly interesting particularly when you
realise the connection to Yabber clients for Gtalk.

Thanks,
   GerardM
_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: OpenID Support

Johannes Ernst-2
In reply to this post by Brion Vibber
I'm sitting in a workshop on user-controlled identity at Harvard as I  
write this ... and the discussion here resonates a lot.

I'd have to concur with Brion here. Anything related to digital  
identity has technical, but also economical, political and *in  
particular* personal/social impact. For example, recall the  
collective outcry when Yahoo made the -- technically, economically --  
reasonable argument that Flickr users should use their Yahoo account  
to identify themselves to Flickr in the future.

We are not just talking about identifiers here in a technical sense,  
but about a public Persona that identifies the registered mediawiki  
user as a member of a particular "tribe" ... and we techies should  
never try to get between a person and their tribe, because we'll  
always lose ;-)

On Feb 7, 2006, at 19:34, Brion Vibber wrote:

> Brion Vibber wrote:
>> Evan Prodromou wrote:
>>> Really? I'd think that an Interwiki-namespaced username would  
>>> solve that
>>> problem nicely,
>>
>> No, it solves a different problem.
>>
>>> without requiring a big name-clash-fixing step when the
>>
>> That step is inherent in the requirement of a single namespace.
>
> Sorry to be short about this, but we've had enough fights over this  
> issue in the
> past, and as far as I'm concerned the decisions have been made.  
> I'll get flak no
> matter whether I implement the previously-agreed conversion  
> specification or
> something new, so it's lose-lose for me personally. ;)
>
>
> However to summarize, my view on why a single namespace is Good  
> includes a few
> points:
>
>
> * Encourages community cohesiveness across diverse projects:
>
> Many people are active in multiple projects. While they may be  
> numerically a
> minority, these are among the most active contributors, and thus  
> key to the
> projects' success.
>
> One of the things making it hard now to pop over to another  
> projects is that you
> have to register another account; if you instead have a "foreign"  
> account saying
> you're from another web site, this makes you stand out even more as  
> not belonging.
>
> We want the Wikimedia projects to be able to share people and  
> content easily and
> freely. Being an equal user on every project helps break down those  
> borders,
> where having a freakish foreign username would build them up.
>
>
> * Related to that, we have projects specifically for sharing:
>
> Commons and Meta are important. They should be as closely  
> integrated as possible
> to make media content sharing and community-wide issue discussion  
> as easy to
> dive into as possible. Being given a foreign name when you go there  
> makes it a
> foreign place instead of just a special part of the wiki.
>
>
> * Avoids the 'tld problem':
>
> Even though we all know top-level domains in the DNS system are  
> supposed to be
> disambiguators, everybody ends up trying to register their name in  
> every
> possible domain -- if you don't, the others mostly go to squatters,  
> SEO scum,
> and phishers who have no legitimate claim to the name.
>
> The disambiguators are unsatisfying, and aren't necessarily enough  
> to pass
> casual inspection -- especially when you legitimately have a  
> hundred accounts
> named 'Angela' or 'Brion Vibber' which _mostly_ are the same person  
> already.
>
> It makes more sense in my view to recognize that the disambiguators  
> don't work
> that well when they're needed and make more trouble when they're  
> not, and just
> jump straight to a single namespace.
>
>
> When we add OpenID/whatever in the future later, that'll allow  
> identification
> from 'foreign' sites in a limited way which is useful. (Eg a post  
> on the
> LiveJournal article claiming to be from Brad Fitzpatrick could be  
> verified as
> really having from from Brad's LJ account, and a post on LJ about  
> Wikipedia
> claiming to be from Jimbo Wales could be verified as really coming  
> from Jimmy's
> Wikimedia account.) But that works because it's *all about* the  
> boundary, and
> drawing attention to the foreign boundary is what you want in that  
> case. That's
> not, I think, what we want inside Wikimedia.
>
> -- brion vibber (brion @ pobox.com)
>
> _______________________________________________
> Wikitech-l mailing list
> [hidden email]
> http://mail.wikipedia.org/mailman/listinfo/wikitech-l
Johannes Ernst
NetMesh Inc.

  http://netmesh.info/jernst


_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: OpenID Support

Josh Hoyt-2
On 2/9/06, Johannes Ernst <[hidden email]> wrote:
> I'd have to concur with Brion here. Anything related to digital
> identity has technical, but also economical, political and *in
> particular* personal/social impact.

I think there is general agreement on this point. As I stated earlier,
my interest in helping with applying the technology of OpenID to
Wikimedia's single-sign-on problem is independent of the decisions
that the organization makes about the user experience. To re-iterate,
I believe that OpenID is an appropriate technology regardless of what
namespace or identifier policies are in place.

Essentially, OpenID can be used without URLs as identifiers for
Wikimedia sites because those sites will already know where to go to
complete the authentication.

> We are not just talking about identifiers here in a technical sense,
> but about a public Persona that identifies the registered mediawiki
> user as a member of a particular "tribe" ... and we techies should
> never try to get between a person and their tribe, because we'll
> always lose ;-)

This is an interesting statement, because as far as I can tell, the
decision of the Wikimedia foundation has been to unify the namespaces
(thus getting "between a person and their tribe"). I note this not to
criticize the decision, but just as further evidence that the issues
are complex.

Josh
_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: OpenID Support

Evan Prodromou-3
In reply to this post by Brion Vibber
On Tue, 2006-07-02 at 18:37 -0800, Brion Vibber wrote:

> > without requiring a big name-clash-fixing step when the
>
> That step is inherent in the requirement of a single namespace.

It is (at least, now it is), which is why I think forcing a single
namespace is a mistake. Setting up a single namespace when multiple
wikis first started would have been great; right now, as you pointed
out, there are 600+ user namespaces, and a big untangling step is going
to be necessary to unify them.

I realize you think it will bring the entire Wikimedia community closer
together, but I'm dubious. Wikimedia projects have always had a very
wide berth as far as policy is concerned; my guess is that this top-down
social engineering effort is going to be poorly accepted. Most people's
first and perhaps only experience with the feature will be finding out
that their username has been taken away.

It also seems unfair that the people who benefit from the minor
convenience of not using a prefix won't actually be paying the price for
unifying the user namespace. I don't think people are going to like
losing their user names just so that I can feel like part of the gang by
logging in as Evan instead of meta:Evan on Ossetian Wikibooks. For my
part, I don't want to pretend that I'm just another user on Javanese
Wikisource or Kashubian Wiktionary; I'm simply and utterly not. I think
it's a benefit for people to know at a glance what language I normally
speak and what project(s) I'm most familiar with. It also makes it
significantly easier for other users to check my edits on my "main" wiki
and evaluate what kind of contributor I am.

I think loosely-coupled authentication with namespace-prefixed IDs is a
very realistic reflection of the current social environment in
Wikimedia. I think the trade-offs (prefixed user names vs. people losing
their accounts) are waaaay in favor of prefixes. I think that OpenID
single-sign-on would be a very, very big improvement for people who have
a lot of accounts. I also think that users who want to associate with
Wikimedia as a whole could easily use their user names from meta: or
commons:. And I think that single-sign-on techniques that require Web
servers to have database access to a global account DB (or -- nightmare!
-- to 600+ local account dbs) unnecessarily constrain server deployment
and will probably negatively impact performance.

Anyways, thanks for hearing me out. I know you've thought this through a
lot, and I realize you've already started cutting code on the
one-big-namespace feature. A lesser programmer wouldn't entertain other
options; I'm glad you took the time to respond.

~ESP

--
Evan Prodromou <[hidden email]>
Wikitravel (http://wikitravel.org/) -- the free, complete, up-to-date
and reliable world-wide travel guide
_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: OpenID Support

Erik Moeller-2
Evan Prodromou:
> I realize you think it will bring the entire Wikimedia community closer
> together, but I'm dubious. Wikimedia projects have always had a very
> wide berth as far as policy is concerned; my guess is that this top-down
> social engineering effort is going to be poorly accepted. Most people's
> first and perhaps only experience with the feature will be finding out
> that their username has been taken away.

The current specs as well as the implementation strategy are aimed at
minimizing conflict and harm to the community as much as possible. More
active users will receive preference in conflict resolution; there will
be at least one site-level announcement of the transition and its
implications. We will use e-mail addresses where available to send out
transition-related information.

The number of active users who will have to change their username will
hopefully be small. The number of users who have accounts on multiple
projects is higher than you would think -- Commons alone has >25,000
registered users, almost all of which can be assumed to have an account
on the other projects. It's true that me being active on the Ossetian
Wikipedia is unlikely, however, _every_ user is a potentially useful
contributor to other projects in their language, particularly Commons
and Meta.

With a single namespace, it will still be possible to see that users are
new to a project because they don't have userpages yet. You will still
get a nice welcome message, and you will still be treated in the ways of
the wiki. We might eventually make it easier to see existing userpages
of a user across projects, but preserving the "newness" aspect is
important for community dynamics.

 From a usability point of view, the constant hassle of having to deal
with prefixes in many different places of the UI would be a major
drawback. These prefixes would invite false conclusions ("Aha, a
Wikibooks user! Therefore .."), and distract from the idea of a
Wikimedia community.

Not to mention that you want a consistent contributions history. If I
authenticate as Meta:Eloquence tomorrow to Wikipedia, what about my
contributions as EnWikipedia:Eloquence today? Identical users using
different authentication across projects would screw up contribution
histories in a big way and be _really_ confusing.

Being able to look at a user's history across all projects/languages is
a feature that will be really useful for things like Wikimedia-wide
votes, where we want to know exactly what a user has done. Given the
existing contributions histories, and the fact that users might use
different authentication prefixes on different days of the week, this
would be next to impossible without a single namespace.

Every single time I see a username, I would also have to see its prefix
and, if I'm not sure, double-check if they really are the person I think
they are. Every new user would have to understand the concept of
prefixes and how accounts work across multiple projects. It's much less
intuitive than a single namespace.

You could address these problems by offering ways to link accounts, and
only allowing Meta as a prefix, but if you start doing things like that,
you might as well go all the way to complete conflict resolution and a
single namespace.

The cost of multiple namespaces is constant, as opposed to the one-time
cost of transition to a single namespace. Finally, a single Wikimedia
namespace is beneficial to the other OpenID projects as well. If we
support OpenID, all the benefits of uniqueness will be transferred to
projects you authenticate against from Wikimedia. When Jimbo
Wales@Wikimedia comments on your blog, you know that it's His Royal
Highness and not Willy on Wheels.

Erik
_______________________________________________
Wikitech-l mailing list
[hidden email]
http://mail.wikipedia.org/mailman/listinfo/wikitech-l