docs: Issue on page ScyllaDB SSTable - 3.x

[annastuchlik]

I would like to report an issue on page https://docs.scylladb.com/manual/branch-2026.1/architecture/sstable/sstable3/index

Problem

From @avikivity about the table:

All the 3.x notation is misleading.

Suggest a fix

Find a better way to represent the information in the table.

@bhalevy Do you have any suggestions?

[bhalevy]

@bhalevy Do you have any suggestions?

I'm not sure what @avikivity meant exactly that the notation is misleading.
We called it 3.x format since those formats were introduced by Cassandra 3.

The ms format, as well as the planned mt format, are ours though, we there is no point in categorizing them as 3.x format.

Maybe we should just drop https://docs.scylladb.com/manual/branch-2026.1/architecture/sstable/sstable2/index.html
It mentions the m[des] formats which is indeed misleading and wrong. It's supposed to be about the la and ka formats that we support for read, but we stopped supporting them for writes for quite some time.

As for the 3.x formats, we can mention that md and me are compatible with the respective Cassandra sstable formats, and that ms (and later mt) is our extension to the format that adds the Trie index, but drop the whole 3.x concept, as the original Cassandra version that introduce those versions is not very interesting to the user (especially if they moved to newer formats in Cassandra 4 / 5).

[annastuchlik]

Maybe we should just drop https://docs.scylladb.com/manual/branch-2026.1/architecture/sstable/sstable2/index.html
It mentions the m[des] formats which is indeed misleading and wrong. It's supposed to be about the la and ka formats that we support for read, but we stopped supporting them for writes for quite some time.

Yes, this one is misleading. I opened an issue to remove it last week: #30149

As for the 3.x formats, we can mention that md and me are compatible with the respective Cassandra sstable formats, and that ms (and later mt) is our extension to the format that adds the Trie index, but drop the whole 3.x concept, as the original Cassandra version that introduce those versions is not very interesting to the user (especially if they moved to newer formats in Cassandra 4 / 5).

Agreed. I'll do it in a separate, after #30148 is merged (to avoid resolving tons of conflicts).

@bhalevy Thank you!