[MTOS-dev] RSS Support

[Bud Gibson]

Jay:

Remarkably, I've suddenly found myself embroiled in a number of test-driven
endeavors.

Recently, in another email conversation on pronet, this came to light:

As our process over the past few releases has focused on user
> acceptance testing by Chris Hall and the QA team, we've not been
> keeping the unit tests up to date. We intend to fix that, to make the
> unit tests a tool for all developers during our evolution into a
> distributed open source team. Shortest term, we're starting with the
> database driver tests, as that's a point of contention in the Cal
> schedule; you can see my checkins this and last week for the slimmest
> sliver of the improvements we'll be making.


I think the move to test-driven development is right on track, particularly
in this setting.  My ONLY recommendation would be to write tests first such
that they serve as the spec. All new functionality is defined by tests.
Nothing goes out the door without passing all the tests.

The rub starts to come when you decide to radically re-architect a product,
basically what happened between MT3 and 4.  At that stage, you're
potentially rewriting the contract (test suite) for important elements of
the app.

I guess my main hope is that the test suite gets a central enough role
during this relatively stable period post MT4 release, that breakage caused
by new releases will be minimal or non-existent. All things point to this
being the case.

Bud

On Fri, May 2, 2008 at 1:46 PM, Jay Allen <jay at endevver.com> wrote:

> Hi Bud,
>
> Have you ever looked at or checked out any of the dev branches from
> code.sixapart.com?  I ask because I wonder if you've ever looked in the t/
> directory that can only be found in the development repo and is not shipped
> with the app.
>
>
> http://code.sixapart.com/trac/movabletype/browser/branches/release-37/t
>
> If you haven't seen that, that is the actively developed[1] and
> comprehensive test suite that the team uses constantly to make sure that
> their development is on target.  While I won't try to convince you that it's
> completely comprehensive (it's not but no test suite is if the app is in
> motion) or that the core team spends enough time making sure that every
> possible interaction with the software has a test case (they don't), but I
> will say that it's an impressive effort that has been going on for a long
> while and will continue in its progress and improvement.
>
> I would be interested in hearing more from you about this because, as you
> are probably aware, this isn't (or shouldn't be) only the concern of the
> core MT team.  At some point hopefully in the very near future, this will be
> a critical concern of EVERY MTOS developer because as the number of
> simultaneously active developers increases on any platform the test suite
> becomes increasingly more important because it does what no human possibly
> can: To efficiently and definitively tell the group when they've made a
> wrong turn and when they are back on course.
>
> --
> Jay Allen
> Endevver Consulting
> 415-200-6985
>
> [1] -
> http://code.sixapart.com/trac/movabletype/log/branches/release-37/t?action=follow_copy&rev=2189&stop_rev=&mode=follow_copy
>
>
>
> On May 2, 2008, at 6:39 AM, Bud Gibson wrote:
>
> > I think it's going to take some time to get an MT that is less prone to
> > breakage on upgrade.  One thing that should really help is the move to test
> > driven development.  I would think that could really help with maintaining
> > the api.  In some sense, the test suite would define the api, and nothing
> > would make it into release that broke the tests.
> >
> > I think a comprehensive test suite is some time away, hence my sense
> > that stability will take some time.
> >
> > Bud
> >
> > On Fri, May 2, 2008 at 8:47 AM, Jay Allen <jay at endevver.com> wrote:
> > On Wed, Apr 30, 2008 at 2:37 PM, Bud Gibson <fpgibson at gmail.com> wrote:
> > >
> > >  Let me speak to the issue of why I am not a plugin fan. Simply
> > stated, they
> > > are a maintenance headache. They're out of the core and they break on
> > > upgrade. Byrne even noted this in a leaked email last August. I'm just
> > now
> > > diving into MT4.1 for a site upgrade, installing the recaptcha
> > plug-in, and
> > > thinking OMG, will this work when the MT4.1.5 upgrade surfaces?
> >
> > Luckily, much of that pain should be abating and eventually disappear
> > now that us developers have full access to the source code after every
> > single change to it by the core developers.  It used to be that
> > developers only had a few weeks (if that) to not only learn all of the
> > new front-end and API features but to quickly fix their plugins.  This
> > was especially harmful for:
> >
> >   * The most prolific plugin writers (e.g. Arvind) who had many
> > plugins to update,
> >   * "One hit wonder" developers who wrote a single plugin and then
> > drifted away from the community far enough to either not know or not
> > care that there was a new MT version out that required updates to
> > their plugin
> >   * Or the worst case, plugin developers who fit into **both**
> > categories by being extremely prolific in the MT 2.x and early 3.x
> > days before completely dropping off the face of the Earth.  I won't
> > name names because you probably know a few of these.
> >
> > So, it's now time to realize that your old biases, though probably
> > well-founded, are now quite solidly out of date.  This is the
> > beginning of the better times we've all been waiting for for so long.
> >
> > --
> > Jay Allen
> > Endevver Consulting
> > 415-200-6985
> >
> >
> >
> > --
> > Bud Gibson
> > cell: 734-657-4800
> > web: http://michiganinnovators.org
> > Now a certified AdWords Professional
> >
>
>


-- 
Bud Gibson
cell: 734-657-4800
web: http://michiganinnovators.org
Now a certified AdWords Professional
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.sixapart.com/pipermail/mtos-dev/attachments/20080502/0b75c8eb/attachment.html