[Seaside-dev] Re: [Seaside Commits] Seaside 2.9: Seaside-Platform-mls.65.mcz

[Julian Fitzell]

On Thu, Jun 4, 2009 at 11:47 AM, Michael Lucas-Smith
<mlucas-smith at cincom.com> wrote:
> Julian Fitzell wrote:
>>
>> I guess that would be fine as long as there are tests defining the
>> protocol that is expected of the returned FileName object. Without
>> those tests or a WAFileName object (also with tests), a platform would
>> have no idea what they needed to implement or whether their
>> implementation was correct.
>>
>
> Okay, I'll spend some more time on getting this right.
>
> On another point, my publishes of Seaside-Platform do the following:
> 66 - undoes the float change
> 65 - adds filenameNamed:
> 64 - adds the float change
>
> All of these are incorrect, is the procedure to publish a new head 67 that
> would remove filenameNamed: or to rewind back to lr.63?

Erm... we occasionally delete things if we're fairly quick and certain
that nobody has loaded them. We could probably do this in this case
but it's also just as easy to load 63, adopt 66, and then save a new
version.

>> Are you doing this because you want to use Seaside-FileSystem or just
>> because it's there? If the former, go nuts. If the latter, we should
>> check whether anyone wants to keep it - I don't think we were
>> currently very happy with the coverage of the Sport Platform and it's
>> an abstract class with no concrete implementations currently. Maybe we
>> should just move the FileSystem stuff out to another repository or
>> delete it altogether. Philippe?
>>
>>
>
> It's the former. We added WAExternalFileLibrary for Seaside 2.8 in VW and
> people found it really useful. This package provides that same(ish)
> capability. Given that people found it useful in 2.8, I suspect they'll
> continue to find it useful in 2.9.
> The package comes from Seaside29LGPL, presumably because it uses Sport.
>
> I'm happy to continue working on it to get its test coverage up.

Ok, sounds good. It might be better to define Platform protocol that
returns ANSI-standard stuff like streams and so on rather than
FileName objects but I haven't really looked at the package so I'm not
sure what's in there. I'm just concerned about the complexity of
implementing the new platform methods for platforms that do not have a
FileName object compatible with Sport or VW or whatever it is you're
proposing to return.

If you're going to have platform-independent FileSystem code, we'll
also have to rename Seaside-FileSystem to Seaside-Sport-FileSystem I
guess.

I'll wait for Philippe to chime in before saying any more because I
think he actually knows the code.

Julian