[Borgbackup] lzma, 3-2-1 rule

[Thomas Waldmann]

> in the german computer magazine c't, I found an article about borgbackup 
> -very nice. There it is mentioned that the compression method lzma is 
> deprecated and only left for compatibility reasons.

Hmm, IIRC our docs rather say zlib/lzma are "legacy" (but not deprecated)?

Legacy meaning they are older algorithms and maybe not as 
nice/efficient/fast/versatile as zstd (or lz4, if one wants highest 
possible speed).

Anyway, consider zlib and lzma are provided by python stdlib, removing 
them from borg would not give us any real "advantage" (like saving some 
dependency), so guess they are not going away for a long time.

Removing any compression algorithm is problematic anyway as there might 
be users having a ton of data compressed with it (like in your case). So 
the only points in time we could do that is a breaking release (like 
borg2) that needs "transferring" archives from an old repo to a new one 
anyway.

For borg2, the is no plan yet to remove any compression algorithm.

> My concern: my actual backup is running since February every night, and 
> the estimated completion might be in June or July

Oh, wow. O.O

> and the constantly slowing down backup rate of now ~5 
> GB/h compressed/deduplicated data.

You checked that borg does not use more RAM than you have (neither on 
client nor or server side)?