Re: [apps-discuss] JSON Schema considered harmful

[Andrew Newton <[e-mail]>]

On Fri, Sep 21, 2012 at 5:18 AM, Hannes Tschofenig
<[e-mail]> wrote:
>
> I don't think that arguing are against the usage of formal or semi-formal ways to describe protocols. Pseudo code may be a useful way to describe an algorithm, for example. The issue is just that they have to be applied selectively and carefully. There has to be a real benefit.
>
> In the context of this discussion I had been arguing that XML schemas (and Relax NG schemas) had not been useful for the target audience compared to the benefits they claimed to provide. For that reason I argued that we should discourage people from using them since they cause problems for extensibility, specification quality, and readability.

As far as extensibility goes, I think that's an issue more with
Namespaces in XML than with the schema languages. However, there could
be improvement. From a spec quality and readability point of view, I
think this is more true of XML Schema than RelaxNG.

Regarding utility for a target audience, from what I have experienced
the problem is that tools vendors convince programmers that as long as
they drop their XSDs or WSDL in spot X, magic will happen and they
don't have to worry about all that nasty protocol crap (including
talking to the network engineers about firewalls)... that is until
they drop an XML document in a string and the whole thing blows up.
But the issue isn't with XML Schema itself, it is with the
magic-happens-here and computers-will-learn grand architecture in
which XML Schema was drafted.

>
> I like the tool idea that PhB and Tim had brought up. I also think that examples are the most valuable tool for implementers.

+1 on the examples.

Perhaps Phillip could get some time in Atlanta in Apps or maybe get a
room for a BoF to demonstrate his idea. Though tools, including
diagramming languages and schema languages, for writing better
specifications seems like a topic for more than just Apps.

-andy