Do we need Atomic Operations?

[whikloj]

Discussion in the Fedora API specification has begun around whether Atomic Operations should be part of the specification.

See
fcrepo/fcrepo-specification#75 and fcrepo/fcrepo-specification#76

At this point I am starting to wonder (like @ruebot did in IRC) whether we really care about Atomic Operations or should we just handle failures in our code as they happen?

Discuss

[ruebot]

If we care about scaling, which has been one of our founding goals of CLAW, then we cannot push for batch atomic operations.

[acoburn]

Batch ops are in direct conflict with scalability. I can elaborate if necessary.

…

Sent from my iPhone On Mar 18, 2017, at 12:35 PM, Nick Ruest <[email protected]<mailto:[email protected]>> wrote: If we care about scaling, which has been one of our founding goals of CLAW, then we cannot push for batch atomic operations. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#557 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ABNQ2xpc8wWX9d8USMq2q_7qZkei6A0cks5rnAfogaJpZM4Mhdzr>.

[dannylamb]

I'm working under the assumption @acoburn just provided and am specifically avoiding them in the code. So FWIW, from my point of view, they aren't neccessary. Even if a Fedora impl provides them, I won't use them. Better to have good error handling/retry scenarios to handle fails as they happen, because I want islandora to scale.

[ruebot]

@dannylamb can you comment or thumbs up over here: fcrepo/fcrepo-specification#76

[dannylamb]

@ruebot Done

[ruebot]

Resolved with:

• fcrepo/fcrepo-specification@4053b68
• fcrepo/fcrepo-specification-atomic-operations@880816b