A while ago I posted details about the RSS Platform Download Engine. That post focused on downloading of feeds, but did not include additional details on enclosure downloads.
Enclosures are, as most readers know, files that are "attached" to items in an RSS feed. Typically, a publisher will include a reference to a binary file, which an RSS aggregator can optionally download when the feed content is downloaded. The most common example of enclosure use in RSS feeds is for podcasting, where the attached (or "enclosed") files are audio files.
As with feed download, we designed the enclosure download with server and client bandwidth in mind since feed as well as enclosure downloads also happen in the background. Their impact on foreground applications should be limited. Similarly, the impact of large enclosure downloads on servers should be limited.
Let me sketch how the enclosure download process works:
- Every time the feed download engine runs it processes feeds that have the "Automatically Download Enclosures" setting set to true it. If it comes across a new item with an enclosure it adds the URL of the enclosure to a FIFO queue.
- Before the enclosure is added to the queue, the URL is checked with Attachment Execution Service API (AES) to assure the enclosure file type is one of the permitted types. If it's not, the enclosure download is failed (IFeedEnclosure.LastDownloadError = FDE_DOWNLOAD_BLOCKED).
- The first 4 enclosures in the queue are then handed off to the Background Intelligent Transfer Service (BITS). BITS is a background download service that ships in Windows and which enables downloading of files in the background while limiting its affects on network usage. In particular, BITS uses HTTP RANGE requests to download files in chunks. BITS also monitors whether foreground applications (like email or browser) are using the network, and if so, it throttles back its own network usage to limit its impact on those applications.
- Once BITS completes downloading an enclosure, the Download Engine uses AES to save the enclosure to the folder corresponding to the feed. Saving via AES associates zone information with the file. The zone information is used when the file is launched at a later time.
- If there are more enclosures waiting to be downloaded and there are less than 4 enclosure downloads active, the next enclosure is handed off to BITS as in step #3.
If, however, the server of the enclosure does not support HTTP RANGE requests, the Platform Download Engine falls back to downloading the enclosure via a regular HTTP GET request. If this attempt fails as well, then the enclosure download is fails and will not be attempted again automatically.
Note that the enclosure fall-back download (HTTP GET) is size limited to 15MB to limit the impact of denial of service (DoS) attacks against the RSS Platform Download Engine. Since the RSS Platform Download Engine runs in the background, a malicious server could consume all of the client's download bandwidth without the user having any idea. Enclosure download via BITS (HTTP RANGE requests) is less impacted by such an attack and is consequently not size limited.
In other words, if you are an enclosure publisher that wants to serve enclosures larger then 15MB to IE7 users, then you should use HTTP servers that support HTTP RANGE requests. Most popular web servers support HTTP RANGE requests.
It's also worth noting, that when a server does not support HTTP RANGE requests, the RSS Platform Download Engine will issue two requests for each file (the first testing for HTTP RANGE support, and the second to download the file without range support).
For more details on the security measures used to protect applications and users from potentially malicious enclosures, see Miladin's enclosure security post.
I hope that this description of the enclosure download process explains the "multiple-requests" that some publishers have seen, as well the security restrictions associated with enclosure downloads.
-Walter vonKoch
Program Manager
Read the entire article here:
..
