Re: [ucn] more notes on installing the unicorn tool on qa-dev

On 8/3/06, olivier Thereaux <ot@w3.org> wrote:
>
> but it still wasn't, and I remembered that the supported mime types
> are also configured in the tasklist.
>
> I don't know if that's the most usable. Could the list of mime types
> accepted for a given task be computed from the observers:
> - concatenate all mime types supported by the observers of the task
> - if an observer does not support the mime type, skip it if the
> document passing the task has that unsupported mime type.

In fact, it's already the case. The mime types defined in tasks give
priority to EACH observation. The list of mime types supported by the
task is the concatenation of all these mime types. For a given
document, if its mime type is contained in the list, the task can be
performed. Then, if an observation does not support the mime type, it
will be skipped.

An example is when using the "Validation" task (containing two
observations: markup and CSS) on a CSS document: the task can be
performed, but only the CSS observation will be made.

> alternatively, another approach:
> - compute the list of supported mime types for a given task as the
> least common denominator from all its observers.

It is probably not a very good solution as it would invalidate the
example given above.

> I prefer the former solution, but it may have its issues. And perhaps
> there were issues that pushed us to define mime type support both in
> observers and tasks, if so, sorry, I have forgotten, could you remind
> me?

We decided to define mime types in both observers and tasks because we
didn't think priority had something to do with observers. Instead,
this is a problem related to task management, that's the reason why we
chose to put it in the tasks (for each observation) rather than in the
observers. At the very beginning, priority was in the observer's RDF
file, but it did not appeared very logical.
Moreover, it allows to keep observers under control. For example, if
an observer can handle  both HTML and PDF files, we may not want to
check PDF documents through Unicorn as it has hardly nothing to do
with W3C. Duplicating the mime types allows that. An observer says "I
can handle these mime types", Unicorn says "I will only use these
ones".

Jean-Gui

> Thanks,
> --
> olivier
>

Received on Thursday, 3 August 2006 11:55:49 UTC