Merge pull request #96 from microsoftgraph/bugfix/ranges-and-compression

fix: adds content-encoding and range precisions

middleware/CompressionHandler.md

@@ -6,9 +6,11 @@ A middleware component that compresses request bodies and falls back to an uncom
 
 - Add `Content-encoding: gzip` to request headers for requests with a body.
 - Compresses request bodies.
-- Should the handler receive a [415 Unsupported Media Type](https://httpwg.org/specs/rfc7231.html#status.415) response or a `400 Bad Request` response with an OData error containing the following message `Unable to read JSON request payload. Please ensure Content-Type header is set and payload is of valid JSON format.` when the content-type was `application/json`, the handler will retry sending the request without compressing it a maximum of one time before passing on the error.
+- Should the handler receive a [415 Unsupported Media Type](https://httpwg.org/specs/rfc7231.html#status.415) response, the handler will retry sending the request without compressing it a maximum of one time before passing on the error.
 - Compression handler should be installed as a default handler by [GraphClientFactory](../GraphClientFactory.md)
 - The compression handler should accept a `CompressionOptions` object with a property `EnableCompression`. That object can be passed when instantiating the handler, if nothing is passed a default value is created with `EnableCompression: true`. That object can also be passed on a per request base from the fluent-style API by the SDK used. If a value is passed for the request, it takes precedence over the value that was set with the handler. Compression is only performed when the computed value of `EnableCompression` is `true`.
+- MUST NOT compress request bodies when a Content-Range header is present with a "bytes" range unit to avoid "corrupting" the set of requests in the range. (e.g. Content-Range: bytes 0-9/20)
+- MUST NOT compress request bodies when a Content-Encoding header is already present to avoid double compression.
 
 ## Remarks
 

tasks/FileUploadTask.md

@@ -30,6 +30,8 @@ This task aims to provide a fluent and easy to use mechanism for the consumer to
 - The task should be agnostic to the kind of upload being performed so as to support for various fileUpload scenarios e.g. **DriveItem** and **FileAttachment**. An example of the agnostic nature of task is how the task is marked as completed considering different response formats from each API:
   - If the response status is a 201.
   - An additional case for OneDrive is, if the response status is a 200 and the response contains an "id" then the task is complete.
+- MUST NOT compress the individual slices when it uploads them. This is because compressing slices with content-encoding header leads to invalid ranges/content length headers. (e.g. for a 100 bytes file of 10*10 bytes uncompressed slices, the content length would be ~7 bytes)
+- MUST NOT compress the whole file before sending it. This is because it might lead to memory issues in some client contexts (needs compressed result length to [compute the range header](https://www.rfc-editor.org/rfc/rfc9110.html#section-14.1.2-3)), it's also because the content-encoding header applies to individual requests and not to the overall set of requests and the server would try to decompress each slice independently, leading to a corrupted result.
 
 > Note: Outlook and Print API does not allow to update an attachment after it has been uploaded.
 
@@ -39,6 +41,10 @@ Refer to the following documentation for more information:
 - [OneDriveItem](https://docs.microsoft.com/en-us/graph/api/driveitem-createuploadsession?view=graph-rest-1.0&preserve-view=true)
 - [Print API](https://docs.microsoft.com/en-us/graph/upload-data-to-upload-session)
 
+### Appendix on the range header for Microsoft Graph
+
+Microsoft Graph large file upload is inspired off [range response syntax](https://www.rfc-editor.org/rfc/rfc7233#section-4.2), [updated in RFC9110](https://www.rfc-editor.org/rfc/rfc9110.html#section-14.1.2).
+
 ## Large File Upload Session
 
 Json Schema of the upload session object: