Edit ZIM metadata "in-place"

[benoit74]

While ZIM metadata are supposed to be right from the beginning, we regularly face situations where the ZIM produced is ok in term of content, but we have few metadata to edit.

Currently, the process we use consists in recreating the whole ZIM with one-to-one copy of most content and some modified metadata. This is both not very convenient and not

What are the possibilities to edit metadata in-place? If not possible to directly change them since their length might change, wouldn't it be possible to just leave old blobs as they are, create new blobs with updated metadata and update the pointers (TitlePtr?) to the new blobs? This means few bytes will be wasted, but since metadata are very small this would be only very marginally noticeable.

[kelson42]

What are the possibilities to edit metadata in-place? If not possible to directly change them since their length might change, wouldn't it be possible to just leave old blobs as they are, create new blobs with updated metadata and update the pointers (TitlePtr?) to the new blobs? This means few bytes will be wasted, but since metadata are very small this would be only very marginally noticeable.

Any modification of a Metadata content implies at least:

• Changing or add the whole cluster
• Recomputing the checksum at the end
• Changing the ZIM indexes

Yes it is probably possible to have a "zombie" blobs and in a way we could probably have a process which would be quicker than what we have currently (but still very expensive).

But this is not the way forward. The ZIM format is not conceived for that and will never work properly for that.

To my opinion, the way forward is to secure that the future CMS and Zimfarm can be kept in sync properly so each change in the CMS will apply properly to the ZIM at the next time will rebuilt it.

I'm obviously not even talking about properly checking the metadata are OK before building the ZIM.