IRC logs for #trustable for Thursday, 2018-06-21

*** toscalix has joined #trustable06:06
*** toscalix has quit IRC07:57
*** sambishop has joined #trustable08:27
*** qwebirc33419 has joined #trustable08:46
*** qwebirc33419 has quit IRC08:52
*** willbarnard_ has joined #trustable08:54
KinnisonHi all09:20
KinnisonI have been poking at and noticed that the Evidence Tracker doesn't appear to be responsible for t.evidence surrounding t.changes to t.environment or t.requirements09:21
KinnisonIs that intentional?09:21
paulsherwoodwillbarnard_: ^^ ?09:35
willbarnard_for t.requirements it should be responsible09:37
willbarnard_we define t.requirements as a type of t.change09:37
* Kinnison would expect t.evidence to be collected for any and all t.changes09:37
willbarnard_t.environments are a different case09:38
paulsherwoodi think Kinnison's point is they shouldn't be different :)09:38
willbarnard_we consider that t.environments are generated and are ephemeral09:38
* Kinnison refers you to
Kinnisont.evidence MUST be produced for each:09:38
willbarnard_we record evidence of there generation in the evidence tracker09:39
Kinnisonconstruction of a t.environment09:39
willbarnard_yes there is a environment construction evidence type09:39
KinnisonSo there's support for it, but the role listed in the doesn't state the need for it?09:40
KinnisonSounds like needs an update09:40
willbarnard_it is the responsibility of the Orchestrator role to generate the environment construction evidence09:41
KinnisonRight, but doesn't state that the evidence tracker is responsible for aggregation and storage of any evidence other than that for t.software09:42
KinnisonAre you stating that the t.requirements and t.environment evidence is included under that umbrella?09:42
willbarnard_Oh, I understand now what you are saying09:43
KinnisonIf so, a clarification in along the lines of "... which includes, but is not limited to, t.evidence surrounding t.changes to or t.requirements, and t.evidence around the construction of t.environments." would help09:43
willbarnard_Yes, the evidence tracker is responsible for storing all the 6 types of evidence09:43
KinnisonOtherwise that role looks like it only cares about tracking changes to source code for software09:44 is different to t.sourcecode09:44
* Kinnison also notes that the Gate is responsible only for assessing five of the six kinds of t.evidence09:45
Kinnisonwillbarnard_: Okay, that wasn't entirely clear to me09:46
Kinnisonin the top of is defined as being composed of one or more t.changes and having associated t.requirements and t.intents09:46
willbarnard_the wording could perhaps be a little better, I will see if I can reword it09:47
KinnisonIt wasn't clear that is considered to *include* all of t.changes to t.sourcecode and t.changes to t.requirement or t.intents, and the t.evidence around all of that09:47
KinnisonThere's just association rather than direct inclusion09:47
willbarnard_for the role of the gate, the sequence diagram might clarify why it is only the 5 types of evidence09:47
Kinnisonwhich diagram?09:48
Kinnisonwell that's super-confusing09:49 says the gate is responsible for assessing the t.evidence09:50
KinnisonThat diagram suggests the gate is responsible for triggering activity09:50
KinnisonThere's no assessment operations in that diagram at all09:50
willbarnard_The diagram only shows the successful case for brevity09:51
KinnisonThere's no assessment going on in the diagram, this confuses me09:51
KinnisonIs the Gate doing anything other than triggering things?09:51
KinnisonThe Gate is only delivered two pieces of data in that diagram (validation result, and policy handling result)09:52
Kinnisonand does nothing apparent with them09:52
willbarnard_in the unsuccessful case the ci pipeline will fail and terminate09:55
willbarnard_if we want to show those cases we would need a whole set of diagrams09:55
* Kinnison fears he's not being quite clear in his concern09:55
KinnisonThe Gate is responsible for *assessing* t.evidence09:56
KinnisonThat diagram shows only t.evidence being logged09:56
Kinnisonnot being assessed09:56
KinnisonTherefore, while it may be valuable in its own right, it does not tell me why the Gate is responsible only for five of the six kinds of t.evidence09:56
willbarnard_I would be happy to change the term assessing09:57
KinnisonWhat would you change it to?09:57
willbarnard_Do you have a suggestion?09:58
KinnisonNo, because I have no idea what Gate is properly doing09:58
KinnisonI am entirely confused09:58
willbarnard_The final step is change review evidence. The reviewer is responsible for review and approval.10:00
KinnisonThe diagram shows the 'Gate' responding to a change by triggering a series of what I would consider pipeline steps in a CI10:00
willbarnard_exactly10:00 states that the Gate *assesses* evidence10:00
Kinnisonto assess is to examine and to ascribe *value*10:00
KinnisonSo either the diagram needs to show that, or the role description is incorrect10:01
willbarnard_I would be happy to change the term assessing10:01
Kinnisonsynonyms for 'assess' include 'measure' 'evaluate' 'appraise' 'value' 'classify' etc.10:02
KinnisonIs the Gate doing anything like that?10:02
* Kinnison worries that he's either missing something critical, or just not communicating well enough :(10:04
paulsherwooda gate doesn't just assess, it opens/closes iiuc10:14
* Kinnison nods, hence my confusion. In the diagram, 'Gate' is acting as a CI pipeline controller10:14
KinnisonEssentially like gitlab-runner10:15
* Kinnison goes for a walk to see if inspiration hits10:15
persiaWithout certainty on how much semantics may have changed, some time ago "gate" was mapped to
persiaSummary being something that performs the act of automerge based on some set of policy-defined criteria10:19
persia seems to be the more modern URL for that, actually10:22
willbarnard_in the case that the triggered actions return a "fail" result, e.g. the artefact construction, change validation or policy check fail then the Gate will not proceed to the next step10:25
paulsherwoodwillbarnard_: quite... so we need something other than 'assess'10:26
persiaThat suggests semantics have changed, as the earlier semantics did not impose serialised "steps" prior to gate processing (although any imposed gate did serialise state progression through a complex flow).10:27
persia(also, insert "artefact" vs, "artifact" rant related to the silliness of copying poor international decisions in an attempt to differentiate national origin of language)10:28
paulsherwoodwillbarnard_: eek... did i break the build?13:43
willbarnard_paulsherwood: it looks like there is a missing except: master in the ci file13:46
*** willbarnard_ has quit IRC15:46
*** sudonym has joined #trustable15:54
*** weyfonk has joined #trustable15:55
*** weyfonk has quit IRC17:07

Generated by 2.15.3 by Marius Gedminas - find it at!