Closed mundij01 closed 2 years ago
the FHIR server will need to store images base64-encoded (use more storage) or convert them before transmission (use more memory and processing.)
Storage is cheap - like really cheap. Its does take some CPU cycles and mem though to pack it.
the client will need to base64-decode the image, resulting in extra processing on their side
Client side processing shouldn't be a bottleneck here. They'll have plenty of resources to do the unwrapping.
authoring / editing of the XML requires editor capabilities that can handle such large files (my VS Code XML plug-in ran out of memory and I needed to increase some settings.)
The image sizes I would expect wouldn't go way beyond something like 1 meg or so pr. image. What kind of images/sizes did you have in mind @mundij01 ?
Wondering if we can reuse FDA's SPL image rules here. Their validation rules say images must be <1 MB. See page 30 of SPL Implementation Guide section 2.2.3.5 (https://www.fda.gov/media/84201/download).
Would that solve this if we include this rule in our IG?
The question is if we're ready to make such hard constraints already
Would it be more appropriate to use a DocumentReference to transmit this information? The trade-off is that the client will need to makes requests to retrieve the referenced documents/files, but I think the extra requests from the client offset the consequences listed above. Interested in hearing others' opinions. Thanks.
@mundij01 What do you mean with using DocumentReference ? store in a repository images and provide a link to them ? Or include them as attachment DocumentReference.content.attachment within the DocumentReference resource ? (Attachment.data is still a base64 binary...
Currently, it is proposed to use the Binary resource to transmit images associated with the ePI. In a current proof-of concept, we found that high-resolution images were needed for the user interface (consumer of the ePI content.) Base64 encoding the larger images significantly increased the size of the Bundle. Thinking about a production setting, there are several consequences of using the Binary resource:
Would it be more appropriate to use a DocumentReference to transmit this information? The trade-off is that the client will need to makes requests to retrieve the referenced documents/files, but I think the extra requests from the client offset the consequences listed above. Interested in hearing others' opinions. Thanks.