v2 API's listTags() doesn't support paging

[trentm]

The Docker Registry v2 API's endpoint to list tags is paginated: https://docs.docker.com/registry/spec/api/#tags-paginated
This module's listTags doesn't yet support that.

Currently that is resulting in the "listTags" test in "v2.quayio.test.js" to fail, because the repo used for the test (quay.io/coreos/etcd) now has more than 50 tags (which I guess in quay.io's default page size). Here is some instrumented output from that failing test:

$ ./node_modules/.bin/tape test/v2.quayio.test.js
TAP version 13
# v2 quay.io
#   createClient
ok 1 should be truthy
ok 2 should be equal
#   supportsV2
ok 3 null
ok 4 supportsV2
#   ping
ok 5 should be truthy
ok 6 have a response
ok 7 should be equal
ok 8 should be truthy
ok 9 should be equal
#   listTags
XXX res.headers { server: 'nginx/1.10.0 (Ubuntu)',
  date: 'Mon, 31 Oct 2016 17:56:53 GMT',
  'content-type': 'application/json',
  'content-length': '602',
  connection: 'keep-alive',
  link: '<https://quay.io/v2/coreos/etcd/tags/list?next_page=gAAAAABYF4XlRL00UY6caJ4wGdywTryh8c_MpJfbRTEhQ6e-X4D9MnzHxe0V6UbhxqtGJBqjOvo8_LSmCqOccyvjPs8T4rR0Sw%3D%3D&n=50>; rel="next"',
  'x-frame-options': 'DENY',
  'strict-transport-security': 'max-age=63072000; preload',
  'x-request-received': 1477936612677,
  'x-request-processing-time': 1004 }
ok 10 undefined
ok 11 should be truthy
ok 12 should be equal
not ok 13 no "latest" tag in listTags:{"name":"coreos/etcd","tags":["v0.4.6","v0.5.0_alpha.0","v0.5.0_alpha.1","v0.5.0_alpha.2","v0.5.0_alpha.3","test","v0.5.0_alpha.4","v0.5.0_alpha.5","v2.0.0_rc.1","v2.0.0","v2.0.3","v2.0.4","v2.0.5","v0.4.8","v2.0.6","v2.0.7","v2.0.8","v2.0.9","v2.0.10","v2.1.0-alpha.0","v2.0.11","v2.0.12","v2.1.0-alpha.1","v2.0.13","v2.1.0-rc.0","v2.1.1","v2.2.0-alpha.0","v2.2.0-alpha.1","v2.1.2","v2.2.0-rc.0","v2.1.3","v2.2.0","v2.2.1","v2.3.0-alpha.0","v2.2.2","v2.2.3","v2.2.4","v2.2.5","v2.3.0-alpha.1","v2.3.0","v2.3.1","v2.3.2","v2.3.3","v3.0.0-beta.0","v2.3.4","v2.3.5","v2.3.6","v2.3.7","v3.0.0","v3.0.1"]}
  ---
    operator: ok
    expected: true
    actual:   false
    at: next (/Users/trentm/joy/node-docker-registry-client/node_modules/vasync/lib/vasync.js:201:4)
  ...

Note that Link header: link: '<https://quay.io/v2/coreos/etcd/tags/list?next_page=gAAAAABYF4XlRL00UY6caJ4wGdywTryh8c_MpJfbRTEhQ6e-X4D9MnzHxe0V6UbhxqtGJBqjOvo8_LSmCqOccyvjPs8T4rR0Sw%3D%3D&n=50>; rel="next"',

From the spec link above:

Link: 	RFC5988 compliant rel=’next’ with URL to next result set, if available

Example:
Link: <<url>?n=<last n value>&last=<last entry from response>>; rel="next"

So we have an inconsistency between the spec and practice (which has happened before). This might need some poking around to get right.

(Currently AFAICT, Triton doesn't use this listTags endpoint, so I don't have an immediate burning need to implement this.)

Suggestion:

• RegistryClientV2.prototype.listTags = function listTags(cb) change to optionally take a first opts arg. However for compat we can't require it.
• By default listTags change to fetch all tags (following the "Link" url as required).
• Add opts.n to allow setting the page size (though really I don't see a great need for making this configurable).
• Add some option to limit the max number of links that will be followed. Perhaps a max number of entries to receive.
• Given the spec'd "last" behaviour, it would be nice to expose that and then allow the caller to manually page through if they need. However, looking at what quay.io does here, I don't see how to meaningfully start paging through from part way. It would be nice to see an example on Docker Hub or from the Docker registry code/image.

[aaronbeall]

So this has bitten us a few times with our internal tools. We end up with more than 50 tags for an image (happens easily because each feature branch is a new tag) and our internal build tools stop working, everyone is confused until someone who's been around for awhile realizes what's going on and deletes some old tags. It'd be great if this could get addressed. 👍

[idelsink]

I think this issue has been resolved? Like I can get all the tags of a node repository with over 2000+ tags

Update: Nvm it indeed does not support pagination. AWS ECR defaults returns 100 tags and the response header is not followed