staltz
commented
3 years ago I think it would be a lot easier to do the opposite: use ssb-fixtures to generate a bunch of data, extract the follow graph from that, and then use the follow graph to boot up netsim. the individual peers in netsim could "bootstrap" their ~/.ssb folder by partially replicating the origin ssb-fixture dataset (note that this "partial replication" has nothing to do with "secure partial replication", i'm talking about just a way of picking a subset of the dataset for each node in the netsim, so they have a starting point and that starting point isn't the same for all).
thus, the ask: would it be difficult to generate contact (follow) messages in ssb-fixtures according to a passed-in follow graph?
rather difficult. I mean, it's possible, but requires working on ssb-fixtures quite a bit.
the prime difficulty i see is associating the identifiers in the passed-in graph with the feed identifiers generated by ssb-fixtures
this is another sign why it's best to use the feed IDs from ssb-fixtures, because the responsibility of ssb-fixtures is precisely with generating that data to begin with. The important thing about ssb-fixtures and how it does generation is that the follow graph is built with statistical sampling that imitates real world patterns of connection. see https://github.com/ssb-ngi-pointer/ssb-fixtures/blob/main/src/frequencies.ts
if you pass in a follow-graph, it won't have these real world statistical properties. if it would, then it would be doing the same responsibility as ssb-fixtures
cblgh
hi hi!
as you are aware, mr @staltz, we are coworkers atm—and i am currently starting up work on a network simulator as part of the ngi work.
one of the considerations i am dealing with right now is where/how/if to use
ssb-fixturesdata wrt the sim. one concern on my mind: i think i want to allow network simulator users to specify their own network topology, i.e. the hops structure tying together different ssb users. currently, if i was to use the fixtures data, that would be hard; the fixtures would define the network topology instead, and i'm not sure yet if that is what i want.thus, the ask: would it be difficult to generate contact (follow) messages in
ssb-fixturesaccording to a passed-in follow graph? you can choose the format you think is best, e.g.the prime difficulty i see is associating the identifiers in the passed-in graph with the feed identifiers generated by ssb-fixtures; but the specific identifiers in the graph don't matter as much as the relationship between the feeds (and of course knowing how the graph identifiers map to the generated identifiers)