Re: how to refer to a user

On 25 April 2012 00:32, Michiel de Jong <michiel@unhosted.org> wrote:

> Thanks for opening this community group! I have a question right away:
> how should we refer to a user?
>
> I consider 5 ways to do this (this is a cross-post from last week on
> the 'unhosted web apps' mailing list:
>
> human-readable ways to refer to a user:
> -----
> H1- by their user address (user@host)
>

this is fine


> H2- with a non-unique description, e.g. avatar + first name + last name
>

this is fine


>
>
> machine-readable ways to refer to a user:
> ----
> M1- by dereferencing a document about that user, e.g.
> http://host/user.html*
>

unsure what the * is here ...


> M2- by dereferencing an interface to that user, e.g. mailto:user@host*
>

as above


> M3- as the user herself is not "on" the web, simply don't refer to it.
> refer instead to her interfaces (e.g. her email address) and documents
> about her (e.g. her webfinger profile)
>

something like mailto:user@host, you mean?  this is a URI, also find


> M4- the w3c way (this is a bit complicated and inconsistent, but i'll
> explain it below)
>

fine also


>
> I think when we ask a user to identify themselves, it should always be
> done in way H1. If we ask a user to identify another user, it can be
> done in way H1 or H2. If we want to display a reference to a user,
> then it can be done with (a combination of) H1 and H2.
>

ok ... this is a UI choice ... seems reasonable ... are you suggesting that
your UI choice is the only one that should be allowed across the whole web?


>
> I when referring to a user in a machine-readable document, the obvious
> way to do it would be M1, M2 or M3. A machine can treat an email
> address as if it were just a magical thing. A human is probably
> incapable of thinking about an email address without thinking about
> the user behind it, so that's why i put M3 only under the
> machine-readable options and not under the human-readable ones.
>

if you mean a URI then yes, it's a good idea


>
> The w3c way is a variation on M3, with the following modifications:
> - instead of saying that when we say mailto:user@host we are talking
> about an email address, we have to interpret the context. If an email
> address fits into the context, then it refers to the email address. if
> a user fits in the context, then it refers to the user owning the
> email address. If both would fit then it breaks. (this can happen for
> instance if you talk about the colour of a building. you might be
> talking about the colour of the web page that describes the building.
>

it's using a URI but not always mailto:  ... are you saying you'd like
mailto: to be the only URI scheme used to identify a user?


> - all interfaces (mailto:, xmpp:, tel:, etcetera) are allowed, but
> http and https addresses are only allowed if they point to fragments
> of documents, not if they point to whole documents. So if you see a
> URN with no '#' symbol in it, then you can be sure that it refers to
> the document. If there /is/ a '#' in it, then it might refer to either
> the document or to the thing described by the document.
>
> Especially the third special rule seems very random to me.
>
> I find the w3c way of referring to a thing crazy and wrong, and I
> think it's holding back the web. I think a lot of people agree with
> this. M1 and M2 are probably not going to get a lot of adoption, since
> pointers are hard to understand. So we can use M3, which means instead
> of:
>

>  user == 'mailto:user@host*'
>
> we say:
>
>  userContactMethods == ['mailto:user@host']
>
> and:
>
>    userDescriptions == ['http://host/~user/foaf.json#me']
>
> I think this is called the fenomenological approach to metaphysics. It
> avoid incorrect statements like user ==
> 'http://host/~user/foaf.json#me' and also avoids scary dereferencing
> of pointers. This is for machine-readable references to a user. For
> human-readable ones, we just use user addresses, and imprecise things
> like first + last name.
>

Tons of stuff gets adoption.  What stays and what goes is a function of
many things, but one important thing is of what scales.  Web arch has a
good track record of scaling.  Small systems dont have to worry much about
idenitfiers.  But when hetrogeneous systems start to interoperate identifer
act as primary and foreign keys, and sometimes composite keys.  Therefore,
how your system scales is a function of how you name things.

Most systems to date are silos and/or walled gardens.  So they primarily
have to be concerned with interop with themselves.  As the web gets smarter
you'll see more mashups and data linked together.  Identifiers become much
more important at this point.



>
>
> Cheers,
> Michiel
>
>