small CAP: Remove Bucket Shadowing

[Marta Lokhava]

Hi all, 

As we're looking into improvements to the bucket list in stellar-core, we found a simplification (and an improvement to bucket merge performance). This is an "implementation detail" of the bucket list for the most part, but it does affect buckets produced, making it a protocol change. The attached draft outlines the rationale and some performance assessment for this change. Any feedback is welcome! 

Marta

[Nicolas Barry]

Looks good overall.

I think I have questions on the section "Protocol Upgrade Changes" that is a bit light on details:

* more details are needed on "merges need to be restarted", specifically I think you're implying that the inputs for any merge are always available by simply looking at the latest bucket list (so giving some idea of how things work would help)

* more details are needed on how we can mitigate the "picking the right time" (including quantifying requirements around that "right time")

* nit: I don't think we have a concept of "in-memory merges"

Nicolas

--
You received this message because you are subscribed to the Google Groups "Stellar Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to stellar-dev...@googlegroups.com.
To post to this group, send email to stell...@googlegroups.com.
Visit this group at https://groups.google.com/group/stellar-dev.
To view this discussion on the web visit https://groups.google.com/d/msgid/stellar-dev/9fe2d79a-7c46-40ca-a0d9-ba6f07a0308f%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Marta Lokhava]

Good point, I added more details. Corrected "in-memory merges" as well, since what it really means is "not yet promoted merges".