Re: [Summit topic] Let's have chalk talks (part 1/?)

On 25/10/2021 22:20, Ævar Arnfjörð Bjarmason wrote:

On Mon, Oct 25 2021, Philip Oakley wrote:

quoted

On 22/10/2021 08:51, Johannes Schindelin wrote:

quoted

Apparently I have to send this in chunks, to fool the Bayes filter of our
beloved mailing list into doing The Right Thing.

This session was led by Emily Shaffer. Supporting cast: Ævar Arnfjörð
Bjarmason, brian m. carlson, CB Bailey, and Junio Hamano.

Notes:

 1.  What’s a public chalk talk?

     1.  At Google, once a week, the team meets up with no particular topic in
         mind, or a couple topics, very informal

     2.  One person’s turn each week to give an informal talk with a white
         board (not using chalk)

     3.  Topic should be technical and of interest to the presenter

[...]

A 'listener' perspective..

At the summit, the packfile's packing algorithm was mentioned as a
historic 'chalk talk' that's now in the technical documentation.

In the meantime, while chatting to a colleague about the birthday
paradox and its relation to Bloom filters, I realised I didn't
understand what our Bloom filters were trying to do and what they recorded.

I had a look at the code and documentation, but there isn't much there
about our Bloom filter implementation. A chalk talk could later be used
in the same manner as the packfile discussion to show what the filters do?

If there is someone who'd like to talk through what the Bloom filters
are doing in Git then I'd be all ears.

That sounds like a good idea for a topic, not to take away from any of
that discussion, but I believe the bloom filters we have are exclusively
used for path filtering if they exist in the commit-graph.

I sort of thought that something like that was the case

I.e. for:

    git rev-list -- $path

See c525ce95b46 (commit-graph: check all leading directories in changed
path Bloom filters, 2020-07-01) for one use-case and benchmark numbers.

Often it's just managing to get these little linkages between these (to
the uninitiated) disparate ideas and system parts. E.g. the
'exclusively' part, and then the use case.

 The "commit-graph" has a similar conceptual distance from the "do they
mean just the DAG (directed acyclic graph)?" when it comes to the
technical docs.

 Working out where to squeeze in such information can be tricky once the
docs have closed in around their existing layouts.
-- 
Philip