Proliferation of dev docs on search engines confuses newbies

[Dan Loewenherz]

I brought this issue up earlier today at DjangoCon, but the basic issue can be summarized hence:

1. User searches to find info on a specific feature, gets directed to dev documentation.
2. Said feature (on the development version) is backwards incompatible with previous versions of Django.
3. User does not know better, assumes Django has a bug.
4. Invalid bug is filed in trac.

EDIT: rejected proposal removed, see comment 4.

[Aymeric Augustin]

Thanks for the suggestion.

I understand the idea, but I'm not ready to wipe the dev docs from Google's index. People should be able to find information about an upcoming feature by googling its name.

Besides, we take care to mention in which version new feature are added. And closing invalid tickets is cheap. The most troublesome change was the new {% url %], and the flow of tickets eventually stopped.

tl;dr The cost of the solution seems too high to me compared to the magnitude of the problem.

[Dan Loewenherz]

This issue wasn't opened with a specific solution in mind--it's just illuminating that this is a problem. I get if the solution I presented is a bit too drastic, but I think closing this ticket ignores the real issue.

[Aymeric Augustin]

Yes, we can keep this ticket to track other ideas to resolve this problem.

For the record, the original proposal was:

Hopefully, removing the dev documentation pages from search engines will help solve
this issue, since those who want to read the dev docs can just click through to
them by using the version navigation at the bottom of the docs pages.

I think there is a small subset of actual Django users who run their applications
on trunk. I don't have data to back that up but I can't imagine it's a lot.

I talked with Alex earlier today about how to deal with this issue--he also suggested
redirecting users from dev -> 1.4 (or whatever the latest version is). I initially
thought it was a good idea, but I thought about it and realized you would have to at
least add some sort of referrer check. I then thought whether it would be ok to do
something like

    def conditional_documentation_redirect(request):
        if not request.META['REFERER'].startswith("http://docs.djangoproject.com"):
            return HttpResponseRedirect #... and so on

This felt wrong to me because if that sort of check were in place, users would
no longer be able to permalink to dev docs.

PR @ ​https://github.com/django/djangoproject.com/pull/43

[Preston Holmes]

I think if there is a choice when addressing this - we need to make special efforts for the tutorials.

Perhaps a special admonition like from #14464

refs: #16954, #19359

[Aymeric Augustin <aymeric.augustin@…>]

In 248bc06ed8904fff30c648737a97c9a01628686f/djangoproject.com:

Add a sitemap for the docs site.

Current docs have a higher priority than dev docs. Refs #18922.

[Aymeric Augustin <aymeric.augustin@…>]

In 170944d1f6f554a44e7f6d735a55d13531a9b7cf/djangoproject.com:

Added /stable/ that redirects to the default version.

Refs #18922. Unfortunately this isn't referenced anywhere and I don't
know how search engines could pick this URL.

Refs #19783.

[Markus Amalthea Magnuson]

Is it a reasonable request that any version of the documentation other than dev and the latest release should be hidden from search indexes entirely?

Tried searching for "django authentication" on Google now, and it shows dev, 1.5, and 1.4, in that order. I think 1.4 should be hidden to avoid this kind of thing.

[Aymeric Augustin]

1.4 is still supported, it shouldn't be hidden from search engines.

[Markus Amalthea Magnuson]

Just out of curiosity, how are versions 1.3 and below hidden from search index?

(They are all in the sitemap, so obviously search engines see them there, although with low priority.)

[Aymeric Augustin]

They aren't hidden, and depending on what you're searching Google can still return various more or less recent versions of the docs.

Our main problem is that the docs for each new release duplicate content from previous releases, and Google favors the original URL in this case.

Canonicalizing /en/1.5/ (or whatever the current version is) => /en/stable/ may help.

[Markus Amalthea Magnuson]

Does that mean that /en/stable/ should be included in sitemap.xml with priority 1, instead of the latest version (i.e. /en/1.5/). Or should both variants be in there? Or are we talking redirects?

Sorry to hassle you with a lot of questions, but I'm looking to hopefully improve the situation, because I'm annoyed by the proliferation too :)

[Aymeric Augustin]

Eric Holscher just pointed me to ​http://docs.readthedocs.org/en/latest/faq.html#can-i-make-search-engines-only-see-one-version-of-my-docs.

We could use that to point search engines to the stable version.

[Tim Graham]

Moved the last action item to ​https://github.com/django/djangoproject.com/issues/242