On Friday, 20 March 2015 at 07:14:58 UTC, Walter Bright wrote:
On 3/19/2015 10:46 PM, deadalnix wrote:
I let this here. Very interesting and relevant to anyone here.
https://www.youtube.com/watch?v=mDZ-QSLQIB8
It's nearly an hour long. Got a TL;DW ?
Language features should be tested with real users using
scientific validation processes, instead of being blindly added
to a language.
This means using tests groups with different types of
backgrounds, applying the features in several situations,
devising specific set of measurable targets and applying the
whole statistic analysis to the results.
For example, instead of having an online forum which syntax would
be desired for the new feature X, have the whole set of options
implemented and see which of them achieves better results with
the respective study groups.
An example that stuck with me was that languages that follow
Algol/Pascal syntax lead themselves to less bugs, than those that
follow C like syntax.
There are quite a few other examples. Also the mention that as
far as the researcher is aware, only Microsoft is pursuing such
studies for language features. Don Syme is in the audience and
gives an example how they did it for .NET generics.
--
Paulo
