spec: describe descriptors and digests #111

stevvooe · 2016-06-01T23:12:52Z

Add a definition for content descriptors and digests, which are used as
content identifiers in the OCI image specification.

Signed-off-by: Stephen J Day [email protected]

This is a draft of the descriptor specification, which includes the digest calculation.

Before merging:

Add json schema definition
Add descriptor examples to unit test suite

Closes #92

cc @philips @vbatts

wking · 2016-06-01T23:17:40Z

descriptor.md

+A _Content Descriptor_ or _Descriptor_, describes the disposition of targeted content.
+A _Descriptor_ includes the type of content, an independently-verifiable content identifier, known as a "digest" and the byte-size of the raw content.
+
+Descriptors SHOULD be embedded in other formats to securely reference external content.


I think this should be turned around to get “Other formats SHOULD use descriptors to securely reference external content”.

+1 on turning it around as this format will stand on its own in #94

Well, +1 on this change but I see below it says they can be independent

On Wed, Jun 01, 2016 at 04:59:36PM -0700, Brandon Philips wrote:

+Descriptors SHOULD be embedded in other formats to securely reference external content.

Well, +1 on this change but I see below it says they can be independent

We got the turned-around “Other formats SHOULD use descriptors to
securely reference external content.” line, but this is still here.
Can we drop this line?

philips · 2016-06-02T00:05:06Z

LGTM after minor nit cleanups.

stevvooe · 2016-06-02T00:14:56Z

descriptor.md

+Before consuming content targeted by a descriptor, the byte content MUST be verified against the _digest_.
+Heavy processing of before calculating a hash SHOULD be avoided.
+Implementations MAY employ some canonicalization to ensure stable content identifiers.
+


Include mention of size verification to avoid Length Extension Attack.

stevvooe · 2016-06-03T00:59:41Z

Most of the feedback has been serviced. Schema definitions and validation of examples have been added.

Note that the schema layout probably isn't ideal, but we may want to revisit in another PR.

philips · 2016-06-03T04:09:47Z

descriptor.md

+++
+draft = true
+++
+<![end-metadata]-->


what is this stuff?

i'm not sure

It was in the top of the other spec. Figured we were using it.

Was this copied over from docker? This looks like part of Docker's docs build system.

perhaps. It's seems fine to drop.

philips · 2016-06-03T04:12:12Z

LGTM

nit: There is some metadata at the header of this file. I don't know what it is for but otherwise looks great.

philips · 2016-06-03T04:13:13Z

@vbatts do you know how we force a rebuild?

Also, @opencontainers/image-spec-maintainers can we get an LGTM? One more LGTM required to merge this.

wking · 2016-06-03T05:49:17Z

descriptor.md

+
+### Reserved
+
+The following are field keys that MUST NOT be used in descriptors specified in other OCI specifications:


You're restricting your reservation to “other OCI specifications”? I expect you can trust OCI maintainers to PR this spec if they want a new descriptor field. The folks you want to warn off are external implementations and specs.

And because it doesn't hurt to also warn off other OCI specs, I'd use something generic like “MUST NOT be defined outside this specificiation”.

philips · 2016-06-07T23:30:20Z

OK, looks like that worked.

Need an additional LGTM from an @opencontainers/image-spec-maintainers

wking · 2016-06-07T23:32:09Z

On Tue, Jun 07, 2016 at 04:30:06PM -0700, Brandon Philips wrote:

LGTM

Do I have to do this again?

After each push 1.

philips · 2016-06-07T23:35:41Z

@wking ick. I will file an issue with @caniszczyk to see if we can only have it decrement by 1 or something. This makes reviewing this stuff super annoying.

wking · 2016-06-07T23:37:24Z

On Tue, Jun 07, 2016 at 04:35:42PM -0700, Brandon Philips wrote:

This makes reviewing this stuff super annoying.

Just don't bother stamping approval on things until they settle down
;).

philips · 2016-06-08T03:29:12Z

@wking I just don't like the system implied mistrust. Like I trust Stephen to just fix nits after I give an LGTM and not require a full re-review of the PR. There is some middle ground here that more accurately reflects the trust model of a collaborative project.

wking · 2016-06-08T03:33:21Z

On Tue, Jun 07, 2016 at 08:29:13PM -0700, Brandon Philips wrote:

@wking I just don't like the system implied mistrust.

It's you're project, so you can obviously pick the rules that make
sense to you. But “$OTHER_MAINTAINER is probably still fine with this
PR after the changes” seems like something you'd only have to assume
if $OTHER_MAINTAINER wasn't paying enough attention to explicitly say
“yeah, that still LGTM”.

philips · 2016-06-08T23:47:04Z

I think this is ready to merge. Can other @opencontainers/image-spec-maintainers take a look and given an LGTM?

stevvooe · 2016-06-09T01:10:36Z

LGTM

Apparently, we can LGTM ourselves. Feel free to wait for another maintainer to LGTM before merging.

vbatts · 2016-06-09T01:28:12Z

LGTM

There are other APIs described in this specification (e.g. the state JSON format, and the in-flight command-line API [1]), but this string covers the configuration file and referenced objects (e.g. the filesystem at root.path). As additional, backwards compatible features are added to the spec (leading to 1.1, 1.2, etc. releases) and supported by runtimes, those runtimes will *still* stupport 1.0 configs. Once a 2.0 spec is cut, runtimes that only support 2.0 (and nothing in the 1.0 line) will no longer support the 1.0 config. My preferred approach here would be to use JSON-LD [2,3,4] to explicitly document the intended semantics for each field, which would allow us to drop the config-wide version and version each field independently. That would mean a breaking change on a particular field would only break compatibility for folks who were using that field. Unfortunately, I haven't had much luck pushing the consensus in that direction. This commit does not add wording about how the runtime and other consumers should handle an incompatible version. We can address that once the command-line API lands. [1]: opencontainers#513 [2]: opencontainers#371 (comment) [3]: opencontainers/image-spec#111 (comment) [4]: opencontainers#510 (comment)

There are other APIs described in this specification (e.g. the state JSON format, and the in-flight command-line API [1]), but this string covers the configuration file and referenced objects (e.g. the filesystem at root.path). As additional, backwards compatible features are added to the spec (leading to 1.1, 1.2, etc. releases) and supported by runtimes, those runtimes will *still* stupport 1.0 configs. Once a 2.0 spec is cut, runtimes that only support 2.0 (and nothing in the 1.0 line) will no longer support the 1.0 config. My preferred approach here would be to use JSON-LD [2,3,4] to explicitly document the intended semantics for each field, which would allow us to drop the config-wide version and version each field independently. That would mean a breaking change on a particular field would only break compatibility for folks who were using that field. Unfortunately, I haven't had much luck pushing the consensus in that direction. This commit does not add wording about how the runtime and other consumers should handle an incompatible version. We can address that once the command-line API lands. [1]: opencontainers#513 [2]: opencontainers#371 (comment) [3]: opencontainers/image-spec#111 (comment) [4]: opencontainers#510 (comment) Signed-off-by: W. Trevor King <[email protected]>

The Docker link landed in 7b39189 (media-types: show relations between media types, 2016-05-02, opencontainers#55), but we've had a descriptor definition locally since d49d055 (spec: describe descriptors and digests, 2016-06-01, opencontainers#111). Signed-off-by: W. Trevor King <[email protected]>

We're not forbidding other OCI specs from using these fields, we're forbidding all specs (except for future versions of image-spec) from using these fields [1,2]. The wording I'm bringing in here matches what 873b9b6 (Add some text about extensions, 2016-06-26, opencontainers#164) landed for the org.opencontainers.* annotation namespace. My personal preference would be to use JSON-LD to assign explicit semantics to every field (after which you don't have to worry about namespacing the fields themselves) [2], but that would be a larger change ;). [1]: opencontainers#111 (comment) [2]: opencontainers#111 (comment) [3]: opencontainers#111 (comment) And later comments in that sub-thread. Signed-off-by: W. Trevor King <[email protected]>

We're not forbidding other OCI specs from using these fields, we're forbidding all specs (except for future versions of image-spec) from using these fields [1,2]. The wording I'm bringing in here matches what 873b9b6 (Add some text about extensions, 2016-06-26, opencontainers#164) landed for the org.opencontainers.* annotation namespace. My personal preference would be to use JSON-LD to assign explicit semantics to every field (after which you don't have to worry about namespacing the fields themselves) [3], but that would be a larger change ;). [1]: opencontainers#111 (comment) [2]: opencontainers#111 (comment) [3]: opencontainers#111 (comment) And later comments in that sub-thread. Signed-off-by: W. Trevor King <[email protected]>

We're not forbidding other OCI specs from using these fields, we're forbidding all specs (except for future versions of image-spec) from using these fields [1,2]. The wording I'm bringing in here matches what 873b9b64 (Add some text about extensions, 2016-06-26, #164) landed for the org.opencontainers.* annotation namespace. My personal preference would be to use JSON-LD to assign explicit semantics to every field (after which you don't have to worry about namespacing the fields themselves) [3], but that would be a larger change ;). [1]: opencontainers/image-spec#111 (comment) [2]: opencontainers/image-spec#111 (comment) [3]: opencontainers/image-spec#111 (comment) And later comments in that sub-thread. Signed-off-by: W. Trevor King <[email protected]>

stevvooe added this to the v1.0.0-rc milestone Jun 1, 2016

wking reviewed Jun 1, 2016
View reviewed changes

stevvooe mentioned this pull request Jun 2, 2016

[Manifest] Unclear if "should" intended as "SHOULD" #109

Closed

stevvooe reviewed Jun 2, 2016
View reviewed changes

stevvooe force-pushed the descriptors-specification branch 2 times, most recently from 47e94ac to 5bb899a Compare June 3, 2016 00:27

stevvooe changed the title ~~[WIP] spec: describe descriptors and digests~~ spec: describe descriptors and digests Jun 3, 2016

philips reviewed Jun 3, 2016
View reviewed changes

wking reviewed Jun 3, 2016
View reviewed changes

philips mentioned this pull request Jun 7, 2016

descriptor: discuss the reserved fields #125

Closed

philips mentioned this pull request Jun 8, 2016

image-layout: add an initial image layout spec #94

Merged

vbatts merged commit 355b19d into opencontainers:master Jun 9, 2016

stevvooe deleted the descriptors-specification branch June 9, 2016 01:37

wking mentioned this pull request Jun 27, 2016

Add text about extensions opencontainers/runtime-spec#510

Merged

wking mentioned this pull request Jul 26, 2016

config: Clarify ociVersion covering the configuration <-> runtime API opencontainers/runtime-spec#523

Merged

wking mentioned this pull request Sep 18, 2016

media-types: Replace Docker link with internal descriptor link #324

Merged

philips mentioned this pull request Sep 21, 2016

Convert a serialization config JSON to a OCI runtime configuration. opencontainers/image-tools#25

Open

wking mentioned this pull request Sep 29, 2016

descriptor: Rephrase reserved wording #348

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

spec: describe descriptors and digests #111

spec: describe descriptors and digests #111

stevvooe commented Jun 1, 2016 •

edited

Loading

wking Jun 1, 2016

philips Jun 1, 2016

philips Jun 1, 2016

wking Jun 7, 2016

philips commented Jun 2, 2016 •

edited by caniszczyk

Loading

stevvooe Jun 2, 2016

stevvooe commented Jun 3, 2016

philips Jun 3, 2016

vbatts Jun 3, 2016

stevvooe Jun 3, 2016

vbatts Jun 3, 2016

philips commented Jun 3, 2016 •

edited by caniszczyk

Loading

philips commented Jun 3, 2016

wking Jun 3, 2016

wking Jun 3, 2016

philips commented Jun 7, 2016

wking commented Jun 7, 2016

philips commented Jun 7, 2016

wking commented Jun 7, 2016

philips commented Jun 8, 2016

wking commented Jun 8, 2016

philips commented Jun 8, 2016

stevvooe commented Jun 9, 2016 •

edited by caniszczyk

Loading

vbatts commented Jun 9, 2016 •

edited by caniszczyk

Loading


		### Reserved

		The following are field keys that MUST NOT be used in descriptors specified in other OCI specifications:

spec: describe descriptors and digests #111

spec: describe descriptors and digests #111

Conversation

stevvooe commented Jun 1, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

philips commented Jun 2, 2016 • edited by caniszczyk Loading

Choose a reason for hiding this comment

stevvooe commented Jun 3, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

philips commented Jun 3, 2016 • edited by caniszczyk Loading

philips commented Jun 3, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

philips commented Jun 7, 2016

wking commented Jun 7, 2016

philips commented Jun 7, 2016

wking commented Jun 7, 2016

philips commented Jun 8, 2016

wking commented Jun 8, 2016

philips commented Jun 8, 2016

stevvooe commented Jun 9, 2016 • edited by caniszczyk Loading

vbatts commented Jun 9, 2016 • edited by caniszczyk Loading

stevvooe commented Jun 1, 2016 •

edited

Loading

philips commented Jun 2, 2016 •

edited by caniszczyk

Loading

philips commented Jun 3, 2016 •

edited by caniszczyk

Loading

stevvooe commented Jun 9, 2016 •

edited by caniszczyk

Loading

vbatts commented Jun 9, 2016 •

edited by caniszczyk

Loading