Server version 8.1 Wishlist

[dragotin]

This page collects a list of issues that we provide as an input for our server friends for their discussion about the version 8.1 raodmap.

• Ability to deal with highly parallelized requests, in particular parallel chunk upload, see Investigate/test parallel chunking "Internal Server Error" #2724
• ...But also for MKCOL and DELETEs and MOVEs
• Checksum secured file transmission (GET and PUT). The server's backend should calculate checksums along the file upload and provide the result as a reply header, so that the client can double check if the file arrived properly. Same for download. This depends on Checksums to Verify Up- and Download core#12417
• Asynchronous Actions, ie. assembling of a chunked upload, trashbin handling etc., see
  Asynchronous PUT core#12097
• Improved performance for upload of small files: WebDAV extention MPUT for multiple file upload [WIP] core#12001 (in general make individual requests fast so we could increase client side parallelism)
• Performance improvements especially for large shared file sets.
• Check whether receiving the quota is a cheap operation (we were told it is but it seems to sometimes bring our whole sync down)

Also, the "server involved" bug reports in the client repo should be considered: https://github.com/owncloud/client/issues?q=is%3Aopen+is%3Aissue+label%3A%22Server+Involved%22

[rullzer]

• Ability to retrieve users and groups on the server to allow sharing with users and groups from the sharedialog

[dragotin]

• an ownCloud app that enables us to test server side error states in our integration tests. It would do nothing else than returning a http error code when a certain file is accessed through a http request. That would be a great improvement to QA. => Write special debug app to slow down parts of ownCloud smashbox#94

[guruz]

• Two pushes of the same file at about the same time result in no changes in the etag. Two pushes of the same file at about the same time result in no changes in the etag. core#7802

[guruz]

• Make sure server's sqlite3 DB glue sets a busy wait handler so we don't get problems when syncing (with parallel requests) and the server DB is locked. Make sure cursors are closed everywhere / Database is locked core#10566 (comment) Investigate/test parallel chunking "Internal Server Error" #2724 (comment)

GOLD COMMENT ;-)

[dragotin]

@craigpg this is our Wishlist for server 8.1 which you asked for.

[craigpg]

Thx guys. @karlitschek, @MTRichards.

[rullzer]

• Ability to retrieve users and groups on the server to allow sharing with users and groups from the sharedialog

This is basically done in pull request owncloud/core#13587

[DeepDiver1975]

@dragotin lets work on this list everytime we meet in N - okay?

[dragotin]

yes.

[PVince81]

You forgot to mention.

• stable file ids: File ids must stay stable core#13310

[PVince81]

stable file ids is done 8.1, except for the trickiest case when files are changed on remotely on external storages:

• stable file ids when file remotely changed on external storage: externally renaming a file/folder in a files_external storage breaks the contained shares  core#11533

[butonic]

• use all cores when discovering files

[moscicki]

@PVince81: stable files ids are a concept "broken by design" for generalized sync purposes, please refer to this document: https://github.com/cernbox/smashbox/blob/master/protocol/protocol.md#lifecycle-and-semantics-of-fileid-vs-path-vs-etag

[guruz]

@butonic

use all cores when discovering files

What is this about? Is this about the server file cache?

[butonic]

@guruz No, the client seems to only use a single core during the initial sync. It is at 100% for a long time. There seems to be a performance improvement opportunity hidden in there.

[MTRichards]

Adding this to the wishlist as it is a dependency for checksums: owncloud/core#11811

[PVince81]

Check whether receiving the quota is a cheap operation (we were told it is but it seems to sometimes bring our whole sync down)

It is only cheap if the used space is already known by the server.
In some cases when files/folders weren't scanned yet it might trigger a scan, which might in some cases take a long time to compute the total size.

Maybe we should find another way for calculating the quota in a much quicker manner (single SQL) when requested that way.

[PVince81]

with "quota" I mean used space.

[guruz]

Simpler version of the checksum proposal: owncloud/core#18716

[MorrisJobke]

Two pushes of the same file at about the same time result in no changes in the etag. owncloud/core#7802

This should be fixed meanwhile with high level file locking (optional in 8.1 and enabled by default in 8.2)

[dragotin]

Closing in favour of #4558