Proxy generated NQ projections through Yarp

[donaldgray]

Once generated named queries are stored in S3. The DLCS currently supports projecting to both PDF and Zip files.

Rather than streaming these 'manually' they should be proxied from S3 via Yarp, similar to how AV and Files are streamed. There is a chunk of logic in handling these requets to check if a control-file exists, if it's stale, if it's in-progress etc. We'd need to think about how best to handle these in a Yarp-pipeline.

[tomcrane]

To be clear, NQ projections that are purely data - like a Manifest - are always dynamically generated; NQ projections that involve putting asset derivatives into some file container (pdf, zip, etc) are always cached in S3.

Is there a bare-minimum dynamic query that could act as a checksum source for the cached "derivative" projection? And be stored in the control file?

[donaldgray]

Is there a bare-minimum dynamic query that could act as a checksum source for the cached "derivative" projection? And be stored in the control file?

Possibly. The NamedQueryRepository parses the NQ request and generates an IQueryable<Asset> which is only materialised if required - we could do something with this to get a checksum, although it would involve hitting the DB more than it currently does.