Železná päsť

Broken development workflow

Due in part to the retarded behavior of the wiktionary administrators, I tried to initiate correspondence through e-mail with other participants in the wikimedias, however they only made the recommendation that I should discuss the issue publicly, so here I will discuss the issue on my own talk page.

So, I thought about some potential improvements to the development workflow for modules instead of using the "sandbox" feature which seems to be somewhat insular and it isn't exactly complementary to the collaborative nature of the wikimedias. Let this be an open letter to the wikimedias, and let them make an authoritative decision about whether the recommendation is worthy of further attention and consideration; the administrators of wiktionary need not attend.

Solution 1: Improvements to lua modules

The most obvious solution is that edits should be annotated with a "test" tag if a developer decides that it shouldn't be published as the default module for all articles on the main website; but instead, the "test" edit would be available to developers who are viewing the website through an experimental "test" mode, which may be considered a universal variant of sandboxes. When developers of a module make a "test" edit, then they can enable the "test" sandbox in order to view articles that have been generated with the "test" edit that they have selected in the module history, which isn't too dissimilar to the preferences feature where users can view the website with a different theme (e.g. monobook) before selecting it as the default. In effect, when another user is unreasonably triggered by module errors and they don't want to contribute to fixing the code and instead outright revert the edits that people have made, they can annotate the edits with a "test" tag. The articles on the main website should be generated from the most recent edit without a "test" tag, and then the "test" tag can be discarded if developers determine that there are no significant errors in the module.

Solution 2: Deprecation of lua modules

Although on the other hand, the way that lua modules are used to implement inflections on wiktionary is actually extraordinarily dumb, and they should at most be responsible for instrumenting database queries and constructing templates. Developers could instead upload a sqlite database that has been preloaded with inflections, pronunciations, syllabification and etymology for entries, which is obviously better than hardcoding it directly into each article on the website because the database would be considered a central authoritative source, unlike content on articles which is completely arbitrary. For examples, I noticed that common software such as aspell can generate inflections for terms in its dictionaries using aspell -l sk expand so obviously there is no need to reimplement the functionality in lua modules on wiktionary since they can be trivially inserted into a sqlite database and extracted therefrom.

In more explicit terms, there should be a generic template for inflection tables and a template that invokes the lua module which consequently queries a sqlite database before instantiating a template that produces wiki-markup code to be embedded into the article. For obvious reasons, the database would be opened in read-only mode because, although I don't know about the administrators on the wikimedias, but I sure don't trust any contributors to insert, delete or alter any rows in the database when they are writing modules.

Additional error categories

So, I have been thinking again recently, which tends to be a bad omen for others, but nonetheless, I realised that there should be error categories on wiktionary and wikipedia for any articles that contain spelling mistakes and otherwise defective grammar. Presumably, such a feature can be implemented easily with an aspell module, which could potentially be available through lua rocks; whenever an article is being scanned for template expansion, the serverside parser can append a category such as "articles with spelling mistakes" when the spelling module finds invalid words that aren't in its dictionary, and then the same administrators who get triggered by module errors can immediately revert edits without exerting any effort to find and correct spelling mistakes.

This is a perfectly reasonable idea as it promotes contributions to the aspell project which is currently looking for co-maintainers as per the GNU website; contributors to all languages of wiktionary and wikipedia can also make contributions to the aspell project to ensure that its dictionaries are as extensive as the natural living language requires them to be, so that articles won't be erroneously subjected to the category of spelling mistakes. It also reinforces the proper use of extant templates on wiktionary and wikipedia to specify the language of the text in an article. When articles in english need to have excerpts from other languages, the spelling module can scrutinise the excerpt in the language that the template has specified, and if the article has excerpts from other languages without a template, then contributors will simply need to insert a template when the article ends up in the category of spelling mistakes.