Summary
A small number of document origins on the AMP Cache will change to a new origin to improve readability of AMP Cache origns.
Background
See AMP Cache URL Format, specifically the section "Domain Name Prefix" for the current AMP Cache origin mapping.
When a document is served on the AMP cache, the publisher's origin is mapped to a new AMP Cache origin, such as example-com.cdn.ampproject.org. The mapping algorithm ensures unique AMP Cache origins, while trying to make the origin as human readable as possible.
Publisher Origins containing a "-" (hyphen) in byte position 3 currently result in an AMP Cache origin using the fallback algorithm described in the above document. For example, the publisher origin:
en-us.example.com
currently maps to the fallback AMP Cache origin of:
covfmowbjemnekovz6xlilc56old3te5kbdukxjd6ypb7buewh5a.cdn.ampproject.org
This occurs because the standard mapping algorithm would otherwise produce:
en--us-example-com.cdn.ampproject.org
Which is an invalid hostname due to a "-" hyphen in byte position 3 and 4. See RFC5891 4.2.3.1:
The Unicode string MUST NOT contain "--" (two consecutive hyphens) in the
third and fourth character positions and MUST NOT start or end with a "-"
(hyphen).
The fallback AMP Cache origin is also used in cases of very long publisher origins that would cause the standard mapping algorithm to produce a hostname component longer than is legally allowed.
Design
Changes to the Origin mapping are disruptive to publishers who may rely on specific AMP Cache origins when authorizing CORS requests from the AMP Cache. Therefore this design only modifies the AMP Cache origin mapping for publisher origins which currently map to a fallback AMP Cache origin due to the position 3 "-" (hyphen) case described in the Background.
The proposed design modifies the Basic Algorithm to add a new step 4 between current steps 3 and 4:
-
If the output of step 3 has a "-" (hyphen) at both positions 3 and 4, then to the output of step 3, add a prefix of "0-" and add a suffix of "-0".
In the example above, this would result in:
0-en--us-example-com-0.cdn.ampproject.org
Domains cannot start with a hyphen, so the "0-" prefix will push a non-hyphen byte into position 3 in all cases.
The "0-" prefix by itself could collide with another publisher domain, in this example: 0.en-us.example.com. Therefore, when the prefix is added in step 4, the "-0" suffix is also added. No domain names end in ".0", so this results in a unique AMP Cache origin.
Release timing recommendation
After this change is approved, the AMP Cache would begin accepting requests using either the previous algorithm or the new algorithm as soon as could be implemented.
The documentation would be updated as soon as could be implemented.
An announcement would then be sent to amphtml-announce regarding the change, to give advanced warning in the event that any publisher is relying on a specific fallback AMP Cache origin in CORS processing.
4 weeks after that, clients such as Google Search could begin sending traffic to the new AMP Cache Origins and the AMP Cache could begin rejecting the old origins by redirecting to the new one.
/cc @ampproject/wg-approvers
Summary
A small number of document origins on the AMP Cache will change to a new origin to improve readability of AMP Cache origns.
Background
See AMP Cache URL Format, specifically the section "Domain Name Prefix" for the current AMP Cache origin mapping.
When a document is served on the AMP cache, the publisher's origin is mapped to a new AMP Cache origin, such as
example-com.cdn.ampproject.org. The mapping algorithm ensures unique AMP Cache origins, while trying to make the origin as human readable as possible.Publisher Origins containing a "-" (hyphen) in byte position 3 currently result in an AMP Cache origin using the fallback algorithm described in the above document. For example, the publisher origin:
en-us.example.comcurrently maps to the fallback AMP Cache origin of:
covfmowbjemnekovz6xlilc56old3te5kbdukxjd6ypb7buewh5a.cdn.ampproject.orgThis occurs because the standard mapping algorithm would otherwise produce:
en--us-example-com.cdn.ampproject.orgWhich is an invalid hostname due to a "-" hyphen in byte position 3 and 4. See RFC5891 4.2.3.1:
The fallback AMP Cache origin is also used in cases of very long publisher origins that would cause the standard mapping algorithm to produce a hostname component longer than is legally allowed.
Design
Changes to the Origin mapping are disruptive to publishers who may rely on specific AMP Cache origins when authorizing CORS requests from the AMP Cache. Therefore this design only modifies the AMP Cache origin mapping for publisher origins which currently map to a fallback AMP Cache origin due to the position 3 "-" (hyphen) case described in the Background.
The proposed design modifies the Basic Algorithm to add a new step 4 between current steps 3 and 4:
In the example above, this would result in:
0-en--us-example-com-0.cdn.ampproject.orgDomains cannot start with a hyphen, so the "0-" prefix will push a non-hyphen byte into position 3 in all cases.
The "0-" prefix by itself could collide with another publisher domain, in this example:
0.en-us.example.com. Therefore, when the prefix is added in step 4, the "-0" suffix is also added. No domain names end in ".0", so this results in a unique AMP Cache origin.Release timing recommendation
After this change is approved, the AMP Cache would begin accepting requests using either the previous algorithm or the new algorithm as soon as could be implemented.
The documentation would be updated as soon as could be implemented.
An announcement would then be sent to amphtml-announce regarding the change, to give advanced warning in the event that any publisher is relying on a specific fallback AMP Cache origin in CORS processing.
4 weeks after that, clients such as Google Search could begin sending traffic to the new AMP Cache Origins and the AMP Cache could begin rejecting the old origins by redirecting to the new one.
/cc @ampproject/wg-approvers