RE: [Freecats-Dev] Yves C. on the API

[Henri Chorand]

Hi,

> I was asked by Henri to look at the API and offer comments.
>
> There is not much to say at this stage. I am a hands-on engineer,
> my limitation being, I reflect on the shopfloor than on the drawing
> table.
>
> My only remark at this stage would be on the "output" level, ie,
> what is being returned by the API. I do not see error messages,
> or any convention thereof, so I believe that this should be
> implemented. For example, the API call for creating a TM should
> make provisions for a possible failure, providing error messages:
>
> > > > >
> 6.1.7 TM creation (by administrator)
>
> Parameters:
>
> INPUT admin ID, admin password, TM name, TM description,
> source language, target language, sub-project values list, category
> values list
>
> OUTPUT return code
>
> > > > >
>
> if the return code is an error, there should be a string with for
> example "Disk full" or "This file name already used" or etc.

No problemo here.

I merely inserted a return code as return parameter in all functions because
I believe the client could handle the display of error messages based on its
own current interface language. For instance, error code "0003" would
display as "Espace disque insuffisant" in French, "Disk full" in English,
etc. Error string tables would be located on the clients, the user choosing
the languages he/she wishes to use at setup.

That way, the user can pick any (available) language and the server won't
care and won't need to manage this in detail.

That said, we can of course add other parameters (placeable strings) if need
be, in order to display the user name, TM name, server name, etc. when the
client can't determine such info by itself like a big boy.

One last remark here, when the time of clients will come: we should display
each error code before the corresponding message. Experience shows that lots
of users are very "creative" when re-typing error messages, error codes are
much more handy as far as user assistance is concerned.


There is another aspect on which I would appreciate to know your thoughts
more in detail: indexing rules.

I still like my first idea, designing a set of rules that can suit one or
more natural languages (e.g. whole words plus all substrings starting from
the first character, with a minimum length of 3, for French and possibly
other Latin languages). Of course, these will be refined.


Anyway, you may have different ideas, and you may also want to share some of
your expertise about the implementation of (fast) indexes. Let us know if
you think we are missing something.


> Otherwise, not much to say. The enterprise is enormous. It's a
> considerable effort!

Yves, at first, one never knows for sure when you are joking.

I seem to remember you told me that building a first (basic or even crude)
version of the TM server was one man-month worth, but I'm afraid you must
have used your own standards here ;-)

Three or four man-months (to be split between Tim, Michael and possibly a
few other ones) seem an attainable goal, especially if Michael & Tim can
quickly select and adapt a rather standard HTTP server component.


Cheers,

Henri