[humanities-dev] Other authors in Open Shakespeare

[Tom Oinn]

On 5 June 2012 13:16, Sam Leon <sam.leon at okfn.org> wrote:
> Hi Iain,
>
> Really interesting question. On the last Open Humanities call we were joined
> by Sarah Moriarty who's very interested in setting up Open Joyce.
>
> Like Rufus, though, I feel it might be sensible to set up more generic
> domain specific instances of the chosen platform like Open Philosophy or
> Open Literature.

I think there are two questions here:

1) What are the benefits of a domain or author (or subject, topic,
colour of original binding...) site?

2) What operations are inherently precluded given a split of this
nature and how can we enable them?

As I see it the reasons to create specific sites include:

* Control and a sense of ownership over a topic or domain. We rely on
communities to create and curate content in these sites, by narrowing
the scope of a site and giving it a distinct identity we create a
banner under which such a community can coalesce. The restricted scope
provides focus and a clear purpose to the site and allows for formal
or informal teams of optimal size (that is to say, relatively small
and strongly internally connected).

* Branding and publicity, if we say 'openshakespeare' people have a
pretty good idea of what to expect when they visit the site. They may
not really understand the 'open' bit but they will have some notion of
who's work they'll be seeing!

The inherent drawback of such an approach is that we're picking a
single axis of differentiation but in reality literature can be
divided in any number of ways (just take a look at your bookshelf or
your local library). If we have open<AUTHOR> for all authors what do
we do when someone wants to create a bibliography or similar of all
works addressing <TOPIC>? Clearly there's no single 'right' way to
split the content at a conceptual level.

This came up in a relatively concrete fashion at the Textus session
hosted by Goldsmiths the other week - one suggestion was to have
annotations which were typed tags to indicate passages in texts which
addressed particular themes. If we had openJoyce, openMilton etc how
would we satisfy a query for 'passages in books written within <DATE
RANGE> annotated as referencing <THEME>'? Where would that interface
live, and how would a user ask such a question across what would be
several different sites?

Fortunately I think we have a sane way out of this particular bind -
the Textus server presents a REST based API for query and text
retrieval, if we were to ensure that this could federate cleanly (i.e.
a query to one Textus instance could query other instances and
aggregate the results) we could have multiple independent sites but
retain the ability to search over multiple instances. As Textus is
built to allow deep links (i.e. link into a particular paragraph or
annotation within a text) the result pages for such searches could
trivially link out to other sites in a fashion transparent to users.
As an aside, this would be a powerful motivation to move to e.g.
openID or similar, otherwise users would have to log in to each site,
breaking the seamless effect we'd like to achieve, but it's not a
killer as no login is required to read texts.

We'd need some kind of central registry of Textus instances to which
queries could be federated but we probably want that anyway, and there
would be technical issues to work out such as performance but I don't
think there's any inherent problem with the approach I've just
described. We'd want UI support to allow users to specify the sources
they wanted to query as well, but other than that the federation
should be pretty much transparent. The default could be to not
federate, but we could make it easy for admins of a Textus instance to
add the ability to distribute queries to other instances, the set of
which would be configurable.

Tom

-- 
Tom Oinn
+44 (0) 20 8123 5142 or Skype ID 'tomoinn'
http://www.crypticsquid.com