XML over HTTP with GZIP - Bandwidth efficiency

In one of my applications, I need to download about 50,000 XML documents, most averaging around 5kb in size. I need to find the most efficient way bandwidth wise to transfer all of it over http, and the XML is gzipped. Which is the most efficient way to download it, using the least bandwidth. Downloading each XML file individually, downloading them in sets, or are they practically the same? I have no idea as I don't really know ho开发者_运维知识库w GZIP would effect this.

Thanks

If you are using HTTP, the primary factor for perceived connection speed is the number of TCP connections established.

Serving it as one master file will help (for only one connection would be established). If the server, supports Accept-Ranges (in other words, the resumption of file downloads) feature, then you can use a 'smarter' client to download the master file in chunks over several connections, and regroup them.

But, if that cannot be done, use a HTTP client that supports HTTP 1.1 Keep-alive. That way, the connection would be re-used over several files.

I found another type of encoding that might be better than GZIP for XML over HTTP, it's called EXI (Efficient XML Interchange). It's currently implemented for Apache web server and probably other web servers will adopt it. Here is a benchmark comparing it to GZIP.