Amphtml: Allow the <base> tag

[tomayac]

The tag is currently disallowed, albeit @cramforce in a tweet said it "could probably be whitelisted pretty easily". The concrete motivation is to support pages that use URL rewriting like below…

RewriteRule ^(\d{4})\/(\d{2})\/(\d{2})\/(.*?)$ /index.php?date=$1-$2-$3

…which rewrites http://example.org/2016/02/25/foobar to http://example.org/index.php?date=2016-02-25 and requires the base URL be set to http://example.org/ for local links to work.

[cramforce]

Filed b/27356051 internally, because this first has to be support in AMP cache serving because it can be made valid.

[Gregable]

Closing this old issue. <base href=/> is now valid. We may extend this in the future to other href values, but this works for the given use case.

[tomayac]

While my first AMP pages have just gone live (e.g., https://cdn.ampproject.org/c/s/blog.tomayac.com/categories/technical), it seems like <base href=/> is not back-translated to the origin URL, as all links currently 404, e.g., https://cdn.ampproject.org/2015/06/24/mediterranea-js-trip-report-151125 that should actually point to https://blog.tomayac.com/2015/06/24/mediterranea-js-trip-report-151125 (a currently invalid AMP page, but irrelevant for the actual issue that <base href="/"> served from the AMP Cache forgets it should actually point to <base href="https://blog.tomayac.com/">, which is not allowed as a value).

(CC: @Gregable @cramforce)

[cramforce]

That seems buggy, indeed. I would have expected the base tag to be removed from the output document and the links to be rewritten.

[jridgewell]

Reopened our internal issue.

[Gregable]

This has been fixed on the cache. New documents or documents that change and are refreshed on the cache should have the correct tags. Still working on reprocessing existing documents.

[sham77]

@tomayac thanks for opening that. It's very common indeed. I have the same issue too, and I use the base tag for the same purpose as you mention.

However attn @Gregable @cramforce it seems this hasn't been fully fixed. Trying to validate my amp pages, I get the error [attribute 'href' in tag 'base' is set to the invalid value]. You can try a test version of it here. http://test1.shoppinginformer.com/amp-tst/stores/Lacoste/363

[jridgewell]

@sham77: We only allow /.

[dandv]

Any progress on allowing actual URLs in the href attribute?