Handle Etag generation in one place #34
Conversation
COUCHDB-2650
|
Good idea! May be also extract attachment etag generation for consistency? |
| Rev = couch_doc:rev_to_str({Start, DiskRev}), | ||
| <<"\"", Rev/binary, "\"">>. | ||
|
|
||
| make_etag() -> |
There was a problem hiding this comment.
a random etag is essentially useless. The only reason cloudant does that (for views) is to preserve the shape of the code relative to couch. Better to omit the etag header entirely if it isn't going to be stable.
There was a problem hiding this comment.
Why it was used in CouchDB then? Not because to overcome some cache issues?
There was a problem hiding this comment.
@rnewson: I don't think changing how views handle etag generation is in the scope of the bug. I could work on it in the context of another bug if you think it needs to be done before 2.0 release. I would approach this by putting #mrst.sig into meta. Then when we get here, we can use #mrst.sig to calculate etag instead of relying on #vacc.etag (which is uuid at this point).
|
Should I call |
|
I think not if only you're going to slowly replace couch_httpd by chttpd. Otherwise, it will make more complicated references between subprojects (they are already quite complicated). |
COUCHDB-2650
COUCHDB-2650
Remove code duplication by using similar function from couch_mrview_show. show_etag from couch_mrview_show considers query parameteres so we don't need to pass `lists:usort(chttpd:qs(Req))` in Extra. COUCHDB-2650
We used to use uuid here. We do proper calculation of Etag now. COUCHDB-2650
COUCHDB-2650
COUCHDB-2650
|
This looks mostly good to me. I'm confused by the changes to add header manipulation functions in 9ab5b83 which don't appear to be used anywhere. Is there a corresponding branch in couch_mrview that's not linked here? |
|
Ah, found it. Reviewing the others as well. |
|
stale PR |
COUCHDB-2650