From: Giulio  Piancastelli <giulio.piancastelli@...>
Date: 2004-11-19T21:23:17+09:00
Subject: Re: (Yet Another?) RSS::Parser test suite

Hi Kouhei,

I'm going to address your concerns in a moment, but apart from the
details of every single test case, I'd like to know your opinion about
the employed technique of testing. Indeed, I'd also like to see an
example from your current test suite: where could I find such a case?

> channel_author.xml:
>   Why does /rss/channel/author exist? RSS 2.0 spec.(*)
>   doesn't say it exists.
>
>   (*) http://blogs.law.harvard.edu/tech/rss
>
> item_enclosure_multiple.xml:
>   Where is the description of multiple enclosures?
>
> rss_namespace_*.xml:
>   Where is the description of RSS with namespace?

In fact, I picked up tests from Mark Pilgrim's Universal Feed Parser
without effectively worring about RSS specifications and such. I
perceive him and his work as somewhat authoritative in the syndication
field.

On the other hand, checking the specification, nowhere is stated that,
for example, there could be multiple enclosure elements in an item;
but, there's nowhere to be explicitly said that there could be only ONE
enclosure element in an item as well. You just suppose that there's a
single enclosure for each item because that's probably the more natural
way of thinking about it.

My attempt is a proposal to leverage on prior art test cases also as a
way to better understand the fragmented world of syndication formats. I
firmly believe that, to use test cases from the Feed Validator and
Universal Feed Parser project, it is not enough to turn to
specifications, especially when those specifications are a sloppy as
the RSS 2.0 one: there's something new to learn, and something old to
re-learn at every step. See my comments about RSS plus Dublin Core and
RSS versions as an example of that.

> rss_version_090.xml:
>   RSS Parser doesn't support RSS 0.9.

OK, so feel free to change the test, or the Ruby driver. As I
previously stated, some errors are due to my being unfamiliar with the
whole model used by RSS::Parser. This is probably one of such cases.
Besides, my driver just tries to do some tests on the RSS 2.0 parser:
if you are going to adopt such testing technique, you'll probably have
to write different drivers for different RSS versions (especially 1.0
and 2.0) or require more modules, depending on how you implemented
things. Anyway, this means that the driver would have to be modified in
the future.

> rss_version_201.xml:
>   I can't understand what means this test.
>
> rss_version_21.xml:
>   ditto.

"There are more RSS version in heaven and earth, Horatio, than are
dreamt of in your philosophy." ;)
See: http://diveintomark.org/archives/2004/02/04/incompatible-rss

> > For example, I've imported 'rss/dublincore' but
> > still seem to be unable to access 'dc_subject' and 'dc_author'
fields
> > in both item and channel.
>
> Where is the specification of RSS 2.0 with Dublin Core? I
> can't support it, if it doesn't exist. Because of
> validation.

There is obviously NO specification of RSS 2.0 with Dublin Core. If it
were, you'll have to provide a similar specification for EVERY SINGLE
NAMESPACE people would like to add elements from, and this is quite
unfeasible, you see.

On the other hand, the RSS 2.0 specification you pointed me to says:
"RSS 2.0 adds that capability [of being extensible], following a simple
rule. A RSS feed may contain elements not described on this page, only
if those elements are defined in a namespace."
These days a number of RSS 2.0 feeds offer Dublin Core elements. Even
if there could be an arguable degree of funkiness within those
practices, that's a fact you definetely have to cope with.
Best Regards,
Giulio Piancastelli.