Radical computing thoughts: Data impedance & automatous computing

Sherry Listgarten heysherry@mindspring.com
Sun, 28 Oct 2001 19:05:02 -0800


> Makes me wonder why subsystems and run-time libs don't have a
> mode intended to be used for
> testing where, with some probability, they return indicating
> an error which could in actuality
> be returned from that same call.

Most QA departments will do negative testing and error seeding, and they
could do this whether or not the components were libraries or web services
or ..., as long as they knew what sorts of errors might occur. (Hey, if the
only error is a fatal error, maybe that's a little easier :) )

But I do think that QA of applications that use web services is going to be
done poorly, if at all, by non-professionals. It requires pretty much
sophistication to do well, including simulation of network delays, etc.
Which again makes me wonder how level this playing-field is.

If you consider publishing (web pages), not that many people run debuggers
or periodic scanners on their web pages to make sure that all links are
valid; they don't try their pages on a variety of browers to make sure the
JavaScript can be executed properly; etc. That is, a *lot* of pages out
their are pretty broken. But documents are resilient -- usually you can use
the part that's not broken.

I don't know if this will be the case with web services, particularly if
every error is a fatal error. If, based on that, one can argue that QA is
more important for web-based applications than for web-based publishing; and
if one can argue that QA is not accessible to the general population; then
one can argue that this playing field is not at all level.

-- Sherry.