On Wed, 9 Nov 2022, Michael Matz via Gcc-patches wrote: > I think everything that's on the web server (even the 2.95 docu) has to be > reachable via a (reasonable) set of clicks from the main index.html. It > doesn't need to be _directly_ from the main index.html, though. > > Also, you simply might want to move the "Current Development" section > to the front if it's currently too cumbersome to reach. > > (E.g. one could move the index of the obsolete versions to a different web > page, make that one un-indexable by google, but leave a trace to that one > from the main index.html).
Although we could use robots.txt to prevent indexing of the old documentation, that's not quite right, in that there's no problem with robots accessing the old documentation for any purpose (indexing, archiving, downloading a local copy of all the docs for an old version, etc.), just with showing it in a search engine in preference to a more recent version. The logically correct way of directing search engines to the preferred version to index is <link rel="canonical">. Suppose we could develop some reasonably reliable way of making automated substitutions in the HTML manuals for old versions. (Make sure the original versions of the files, generated years ago, are backed up first! Indeed, when we want to revise such substitutions, we should go back to the original versions of the files to do so rather than piling one substitution on top of another - i.e. the original versions should be kept somewhere readily available for that purpose, just not directly in the web area.) Then we could substitute such <link rel="canonical"> to point to a more recent release (that has a matching documentation page). And as I mentioned at the Cauldron, I think having human-visible notices on the pages for old versions, linking to the corresponding page for human readers - like there are in old versions of the Python documentation - would be a good idea as well. Maybe it would be hard to find matching pages in an automated way across the Sphinx transition - so all EOL versions before GCC 12 would end up pointing to corresponding pages in GCC 12 documentation and a more general index for later versions, rather than a corresponding page for a later version - but whatever can be done automatically in terms of rel="canonical" for search engines and a prominent notice at the top of each page for human readers still seems like it would be a significant improvement (and something that could be improved further incrementally). -- Joseph S. Myers jos...@codesourcery.com
