Sense for a remote .ome.tiff

[yarikoptic]

I arrived at https://avivator.gehlenborglab.org/ and was excited to try out on our data in dandiarchive. Eg. see https://dandiarchive.org/dandiset/000026/draft/files?location=sub-I38%2Fses-SPIM%2Fmicr&page=1 and there for any relevant file click on blue "i" to get to metadata which would have URLs under

    "contentUrl": [
        "https://api.dandiarchive.org/api/assets/3fec7df4-e31c-4327-979f-330a6c84eaec/download/",
        "https://dandiarchive.s3.amazonaws.com/blobs/7ab/b62/7abb6225-a8a8-46e3-8f54-3643c333c5f1"
    ],

where API one would potentially mint a URL (with mimetype provided in the header IIRC) or just redirect to S3 one below, but would require our CORS policy to be adjusted first. So I have tried a direct url to S3: https://avivator.gehlenborglab.org/?image_url=https://dandiarchive.s3.amazonaws.com/blobs/7ab/b62/7abb6225-a8a8-46e3-8f54-3643c333c5f1 but it errorred out with console showing only attempts to treat it as zarr prefix:

[ilan-gold]

@yarikoptic The issue is the file does not end in ome.tif or ome.tiff so Avivator has no way of knowing if the file is a tiff, and thus assumes it to be a zarr store. If you wanted to contribute some sort of file-sniffing, we would accept a contribution.

[manzt]

I’d rather avoid content sniffing. It opens up a lot of complexity and edge cases. An explicit "format" query parameter, which clearly defines how the URL should be interpreted, feels more robust.

[ilan-gold]

I am also fine with not sniffing. I would be happy with a query parameter instead.

[yarikoptic]

sniffing indeed might be tricky, although fetching 2kb or so and running some file (aka magic) library to determine mimetype would be sufficient for some, but indeed for .nii.gz and other "composites" would require more dances.

yet another solution, inspired by the similar option added to niivue for the same reason, and which would make it readily usable for me, is to just provide filename as yet another option. E.g. here is how it works for niivue for .nii.gz: https://niivue.github.io/tiff-loader/?image=https://dandiarchive.s3.amazonaws.com/blobs/f19/d8a/f19d8a9f-7bd6-4bf9-a643-d4694e7d7343&filename=sub-MR243_sample-brain_acq-DSI_desc-DWImasked.nii.gz (it does fetch 5 volumes so might take a bit)