2011/4/4 Amir E. Aharoni amir.aharoni@mail.huji.ac.il:
For example, in the Hebrew Wikipedia there was a Search and Replace gadget long before the advent of Vector's Search and Replace dialog. It was developed due to popular demand, bottom-up, by a volunteer, and
- here's the scariest part - without any grants. It is still used in
the Hebrew Wikipedia, probably much more often than the Vector thingy, which is still rather useless due to bugs such as 20919 and 22801.
As lovely as bottom-up gadget development is, it also highlights the complexity of our challenge in improving usability: By allowing every community to independently develop improvements to things like the toolbar, we're very much creating a risk of degrading usability over time. After all, if you're complaining about the lack of data and formal testing supporting Vector, what justification is there for the vast majority of user-contributed JS changes, which in many cases have terrible UIs and have no formal or informal user testing or supporting data?
And honestly, Hebrew Wikipedia is a great example of this. Just a year after Vector, the standard edit page that even logged out users see has a whole new row of icons in the "Advanced" section of the toolbar, including some very non-intuitive or just plain ugly design choices which are inconsistent with any of the existing icons. Is there any supporting data for the choices that were made as to what was added to the toolbar?
Of course the answer isn't to prevent gadget development, but I do think we need (as Brion highlighted in the wikitech-l thread) much better support systems, consistently enforced style guides, etc. In addition to better analytics systems, that _should_ ultimately include access to WMF design and user testing resources to validate gadget changes, better standard code and icon libraries that gadgets can use, etc.