1618253 - IPC Design for Project FOG

Status: RESOLVED FIXED

(Toolkit :: Telemetry, task, P1)

Description

[Chris H-C :chutten]

How in the flippity flop are we gonna support multiple processes and process types wanting to access the Glean SDK?

Some things have already been decided (though we can always reverse these decisions if there's reason)

• One Glean SDK to Rule Them All - There will be exactly one Glean SDK per application instance, to simplify storage and keep complexity out of the SDK
• There will be code generation - the existing Telemetry approach of hand-rolling code for each and every new metric type we want to support will not scale to Glean-numbers of metrics types.

Pretty much everything else is up for grabs.

On my list of people to consult once I have a doc with at least a concrete framework in it are:

• nika
• froydnj
  (please feel free to add others (or yourselves) if you're interested)

Comment 1

[Chris H-C :chutten]

Attachment: [FOG] IPC Design document

Please find attached the IPC Design document

Please provide early-draft feedback in comments on the doc. If you don't want to, don't feel you have to.

I am not at this time seeking sign-off (though if you'd like to be a person consulted for signoff, please let me know and I'll request your more formal review in a little while), but instead am interested in first- and second-glance impressions. Things like:

• Is this complete? Could you imagine someone being able to take this and make a working system out of it?
• Is this correct? Did I make factual errors?
• Is this icky? Do you have feelings and opinions about the choices (especially the ones prefixed with "I propose" or "Open Question")?
• <and anything else>

If you'd like to discuss parts of it more synchronously than gdocs' comments allows, please find me on Matrix or Slack. I'm at your service.

Comment 2

[Jan-Erik Rediger [:janerik]]

Comment on attachment 9131161 [details]
[FOG] IPC Design document

Left some feedback and questions. I think we're reaching a good way forward with the document. This will probably bring up some more in-detail questions once we move to the implementation, but some of the high-level parts seem sufficiently solved.

Comment 3

[Chris H-C :chutten]

Thank you everyone for your feedback. It seems as though there's nothing obviously missing (excepting a finer level of detail (which we'll get to when we get to implementation)), and it hangs together.

I'm about to go on a week of PTO. If you haven't yet committed any feedback, feel free to provide (or cancel your f?) during that interval, I just won't be around to respond to it until Monday the 23rd. At which time I'll launch into the proper and final presentation & review round to ensure we're not stepping on anyone's toes when time comes to implement this in late April and May.

Comment 4

[Chris H-C :chutten]

This design can now be considered "done" (a word which here means "active writing and feedback phases have completed and implementation can begin, but feedback continues to be welcome and the implementation will of course change everything").

Thank you everyone for your feedback!

Comment 5

[Chris H-C :chutten]

Weird. For whatever reason clicking on the attachment doesn't work anymore, so here's the doc link itself: https://docs.google.com/document/d/1nzFn9riiOa89c6boKxa0umub03FjToygb7-cZp0M1fU/edit#