Child Item

Dropbox API v2 for Second Life

Update (01/14/2019): Finally back to trying to use this code. A resolution was seemingly made, but not fully implemented. I am able to access the API now, but I still cannot use file/download. I have detailed more of this in my latest post.

Hello. I have happily been a user of yours for last several years, using the Dropbox API from within Second Life. This has given me a rather powerful tool for handling data storage, but your v2 API will not function from withint Second Life any long.

I have been trying to migrate from v1 to v2 ahead of time, and I ran into an error that cannot be fixed within Second Life.

Error in call to API function "files/download": You provided a non-empty HTTP "Content-Type" header ("text/plain; charset=utf-8"). This API function requires that the header be missing or empty.

The problem lies here in Second Life:

http://wiki.secondlife.com/wiki/LlHTTPRequest

You cannot override the Content-Type headed, because it is automatically generated from within Second Life itself. Can you fix this problem? There seems to be little reason that the API could not ignore this content type header if you are doing a download, like it did before.

This prevents me from upgrading to v2, and would break all of my current code in Second Life if it cannot be fixed.

Find more posts tagged with

API

Abuse

Comments

Greg-DB

Dropbox API v2 does expect the Content-Type to match the type of the content in the body, and in the case of content-download endpoints, this should accordingly be missing or empty, as there is no content in the body.

Are you perhaps able to explicitly specify the empty string "" using "HTTP_MIMETYPE" as mentioned in the documentation you linked to?

SephDragoon

It cannot accept an empty string sadly. I can attempt to get them to add that functionality, but I do not know how long that will take. I would still like to pursue both options, since either solution, accepting an empty string or ignoring the MIME type when it is expected to be empty.

SephDragoon

This is the exact error returned when I try it:

[06:06:37 PM] AoW:Reloaded HUD (v0.0.6a) - v2 API Example: Unsupported or unknown Content-Type.

Greg-DB

Thanks for trying that out. I can't make any promises, but I'll send this along to the team as a request to allow this on our side.

SephDragoon

Alright. Thank you so much for working with me. Hopefully it is all handled before v1 goes away.

oz_linden

We are investigating whether or not we can, in a backward-compatible way, make a change to how Second Life sends requests like these, but I would respectfully suggest that the check your API is enforcing does not add value and is not appropriate.

HTTP defines a few different ways for the recipient of a request to determine whether or not there is content in the request body, including Content-Length and Transfer-Encoding. If Content-Type is not a multipart type (which it is not in this case), then one of those other mechanisms is required to determine the size of the body and the type specifies only how it should be interpreted. It is quite possible that some end users (possibly including ours - we have not investigated yet) will find that a proxy somewhere in the path will have added a Content-Type header even though there is no content; nothing in the HTTP specifications prohibits this, just as nothing in the specs requires that you reject a request that has a header that isn't needed to interpret the request.

You've added a check that makes your API incompatible without actually serving any useful purpose; I suggest that you allow and ignore the header if there is no content in the request.

Greg-DB

Thanks for the feedback!

oz_linden

Oh, I forgot to let you know.... I'm the Director of Second Life Engineering, and one of the contributors to the HTTP 1.1 specifications at the IETF.

SephDragoon

So, is there ever going to be anything done about this? We're coming up on the date when v1 can no longer be used, and I have NO solution to this. Anyone?

Greg-DB

I don't have an update on this on our side unfortunately. I'll check in with the team, and follow up here if/when I have any news.

SephDragoon

Okay well, I think I've been more than patient enough. Today was supposed to be the day I wipe my hands of this and just accept you had no interest in fixing the problem. It turns out though that you intend to keep me in limbo for another few months.

So, are you or are you not going to fix my issue? I already had to make the announcement I was shutting down because you guys refused to change this -one- line of code to allow me to stay on. I am not going through 3 more months of limbo wondering if you will or won't. I need to know if my development process is done or not; if I am shutting down, or if you guys actually intend to fix it.

I believe this is entirely reasonable. If you don't think I am important enough to worry about, and you don't intend to fix it, then just say so, and I will move on with my life. If you're refusing to answer because you're trying to save face, I don't appreciate it, nor do my users. I would like an answer for once since the six months ago I asked.

Greg-DB

Thanks for checking in. I don't have an update on this request unfortunately. We generally don't make promises about potential changes one way or another, since plans can change and we don't want to accidentally mislead anyone. That said, I'll check in with the team again to see if I can get a firm answer one way or another for this.

Greg-DB

The team has decided to update the API to allow some Content-Types such as "text/plain; charset=utf-8" on download endpoints. I'll follow up here once that is live.

SephDragoon

That is really good news! Thank you for sharing, and I will give an update about it on my end as well!

Greg-DB

API v2 download endpoints now accept the following Content-Types:

- application/octet-stream
- text/plain
- text/plain; charset=utf-8

Hope this helps!

SephDragoon

Hello. I have finally come back to the platform on another attempt to use the endpoints. Picking up where I left off, I decided to finally try and see if it would work, but upon trying it out, I got the same error as before. I am specifying the MIME type as both "text/plain" and "text/plain; charset=utf-8". Neither one shows any different result.

So far it is spitting back "Unsupported or unknown Content-Type." once again. Maybe something else has change since my original attempt, but are the new supported content types still working, or was something eventually changed?

Edit: Alright, I have finally been able to confirm that the error stems only from file/download. Other API calls seem to be working, but I am unable to get the contents of the .txt files I had on the server previously. The content type is still incorrect when I am receiving a response from the server.

http://wiki.secondlife.com/wiki/Http_response

Status 415 "Unsupported or unknown Content-Type"

The remote server did reply to your request but the Content-Type of the reply (such as XML, JSON, Atom, RSS, PLS) is not recognized by the LL server and so is not passed back to the script. You can assume that 415 means the server heard your request and did reply.

Greg-DB

I just tried this, and the new supported content types are working for me. Here are a few sample calls:

$ curl -X POST https://content.dropboxapi.com/2/files/download \
>     --header "Authorization: Bearer <ACCESS_TOKEN>" \
>     --header "Dropbox-API-Arg: {\"path\": \"/test.txt\"}"
This is the test file data.

$ curl -X POST https://content.dropboxapi.com/2/files/download \
>     --header "Authorization: Bearer <ACCESS_TOKEN>" \
>     --header "Dropbox-API-Arg: {\"path\": \"/test.txt\"}" \
>     --header ""
This is the test file data.


$ curl -X POST https://content.dropboxapi.com/2/files/download \
>     --header "Authorization: Bearer <ACCESS_TOKEN>" \
>     --header "Dropbox-API-Arg: {\"path\": \"/test.txt\"}" \
>     --header "Content-Type: application/octet-stream"
This is the test file data.


$ curl -X POST https://content.dropboxapi.com/2/files/download \
>     --header "Authorization: Bearer <ACCESS_TOKEN>" \
>     --header "Dropbox-API-Arg: {\"path\": \"/test.txt\"}" \
>     --header "Content-Type: text/plain"
This is the test file data.


$ # using the verbose option for more detail:
$ curl -vX POST https://content.dropboxapi.com/2/files/download \
>     --header "Authorization: Bearer <ACCESS_TOKEN>" \
>     --header "Dropbox-API-Arg: {\"path\": \"/test.txt\"}" \
>     --header "Content-Type: text/plain; charset=utf-8"

> POST /2/files/download HTTP/1.1
> Host: content.dropboxapi.com
> User-Agent: curl/7.54.0
> Accept: */*
> Authorization: Bearer <ACCESS_TOKEN>
> Dropbox-API-Arg: {"path": "/test.txt"}
> Content-Type: text/plain; charset=utf-8
>
< HTTP/1.1 200 OK
< Server: nginx
< Date: Tue, 15 Jan 2019 16:25:48 GMT
< Content-Type: application/octet-stream
< Content-Length: 27
< Connection: keep-alive
< accept-ranges: bytes
< etag: W/"73c1c021eccc7"
< pragma: no-cache
< cache-control: no-cache
< original-content-length: 27
< dropbox-api-result: {"name": "test.txt", "path_lower": "/test.txt", "path_display": "/test.txt", "id": "id:25N5ksooX-sAAAAAAAKn0Q", "client_modified": "2019-01-15T16:24:48Z", "server_modified": "2019-01-15T16:24:49Z", "rev": "73c1c021eccc7", "size": 27, "content_hash": "cceac37da5288a6b6716737310c14305c9b16a6e0ff0abce4a306c3de3a59f75"}
< X-Server-Response-Time: 1984
< X-Dropbox-Request-Id: c87ab76814c43ef58c5d284e646fdec7
< X-Robots-Tag: noindex, nofollow, noimageindex
<
* Connection #0 to host content.dropboxapi.com left intact
This is the test file data.

While you can now specify any of those three newly supported Content-Type values (or none) for the request, the response's Content-Type for /2/files/download is not guaranteed to be the same as that of the request; it is always "application/octet-stream" to indicate that it is returning the binary data for the requested file (assuming the call succeeded).

I'm not familiar with your platform, but from the SecondLife documentation you quoted it sounds like the issue now is that "LL server" isn't accepting the the "application/octet-stream" response Content-Type for the response from Dropbox for the /2/files/download call.

SephDragoon

I know how to get the calls to function. The platform is still fairly straight forward. You can say that Linden Labs is at fault for their specifications, yes. However that is the easy way out ignoring that if version 1 of the API was not depracated, this would still be functioning. I wrote this system in v1, and there was no guarantee that it would continue to work, but right now the point remains that the differences between v1 and v2 have broken code for some people to a point where it could not continue after version 1 was deprecated. I had to abandon my project a year and a half ago, and after being told it would work I came back to it. Only, it doesn't work.

Greg-DB

Thanks for the feedback! It's much appreciated, and I've shared it with the team.

As far as specific changes to API v2 are concerned, the original requested change here, to support more Content-Type values for requests to /2/files/download, was implemented and is working. Is there another specific change that would be helpful for API v2? I can't make any promises as to whether or not we'd be able to make any particular change, but I'd be happy to send along another request to do so.

SephDragoon

Well, the request being accepted in other content types only does any good if the response does the same thing. Right now the request is working, but I cannot receive the response since its content type is fixed. One does little good without the other.

OzLinden

I suggest retrying your POST test with an

Accept: text/plain

header

SephDragoon

That I will try as soon as I get home, thank you.

Greg-DB

I recommend looking into getting the LL server updated to accept the "application/octet-stream" Content-Type, as it looks like that's where the current rejection is taking place.

I'll be happy to pass along a request to change something on the Dropbox side too, but I should reiterate I can't guarantee such a change would be implemented. Can you specify exactly what change(s) would be necessary though?

Also, I just checked and unfortunately specifying an "Accept" request header like OzLinden suggested won't change the result in this case.

SephDragoon

Well, there goes that. If some solution like what he suggested could be made though, it would help. Otherwise a way to get plain text responses would be lovely. I can understand if that won't work for most file types, but plain .txt files shouldn't be a problem.

Greg-DB

Ok, thanks! I've sent that feature request along. I'll follow up here with any updates on that, but at the moment I'm not aware of any such planned update.

OzLinden

I'm afraid that there's no way for us to modify Second Life to allow anything but a text-based content type.

I'm not clear on how to test the Dropbox API... is it true that if I upload something as text/plain I can only get it back as application/octet-stream? That doesn't seem very useful...

SephDragoon

It works very well for what it is. It had been my go to method of data storage from second life for a long time. The main advantage is that unlike a php to sql method, you do not need to use an intermediary server. The downside is that all logic then has to be put into the LSL script. I may have found a method to make it work for now, but it is a hack using the get_temporary_link call.

Greg-DB

@OzLinden If you want to test the Dropbox API, I have a few examples in this comment from earlier in the thread that show use of the /2/files/download endpoint. The API v2 Explorer can be useful for testing/prototyping calls like this.

In any case, the /2/files/download endpoint does return "application/octet-stream" as the response's Content-Type for any type of file.

@SephDragoon I'm glad to hear you may have found a workaround!

SephDragoon

I would not consider it a solution, but I am able to get the data back out. Since upload works as is, I can still use drop box, but I will be relying less heavily on it. A two step process for every data lookup isn't ideal.