OK, I buy this solution. I updated HLS accordingly, removing the "set
compression algorithm" service. I put the previous design as alternative A2.
Good thing is that now we have one service less and only 3 alternatives to
Andreas Almroth wrote:
> Hi Rafal,
> I think option c is the best sensible solution. We can't stop the BSM implementation
> and user from using their own URL options. If they use WITH COMPRESSION the user
> explicitly tells mysqld to compress and options in URL may be counter-productive, and no
> concern to underlying storage system. We can't cover for every scenario, and in the end it
> will have to be up to the user to use backup operations properly. I think we will just
> have to write in the documentation on how to use BSM compression vs mysqld compression.
> As at least the zlib code already exist we can re-use it I guess, but I don't know
> for sure where to put it in the actual implementation. But I will figure it out...
> Best regards / Cordialement
> Andreas Almroth
> -----Original Message-----
> From: Rafal.Somla@stripped [mailto:Rafal.Somla@stripped]
> Sent: 22. oktober 2009 14:41
> To: Ingo Strüwing; Andreas Almroth
> Cc: backup@stripped
> Subject: WL#5046 (pluggable storage modules) - how to request encryption
> Hi Again,
> I'm afraid I misunderstood Ingo's suggestion of how to specify
> native/internal compression and did not describe it correctly in HLS
> (alternative A2). Looks like now we have 3 alternatives around:
> a) Compression specified using WITH COMPRESSION clause only. Backup kernel
> negotiates compression with BSM as described in the HLS.
> b) Compression specified using location string only. Backup kernel never
> does compression and don't care about it.
> c) User can request compression either using WITH COMPRESSION clause or in
> the location string. In the former case, the built-in gzip compression is
> used. In the latter, BSM handles compression transparently to the kernel.
> Option a) is what I use in my design proposal. Option b) is described as
> alternative A2. Option c) is not described in HLS right now.
> To simplify, I'm thinking about giving up option b) now and considering only
> options a) and c) in further discussions. My question is to Andreas - are
> you OK with that or want to keep option b) as well?