Suppose that you wanted to preserve some textual library and ensure that it remained available. Perhaps you could ensure that by enticing people to comment and contribute, but requiring them to publicly self-host a replica in order to do so.
In particular, you could do this by requiring them to comment by adding the comment in their own local fork of the library and sending a pull request. They could self-host privately, or read without commenting, and remain private; but in order to comment, they would need to host a copy themselves and keep it online long enough for the author to pull from it.
This would ensure that many copies of this library would exist.
Ideally, downloading a replica of the library wouldn’t take any more bandwidth than downloading a regular web page. But how much is that?
Right now, 2015-10-15, loading without cache:
| what | how much | reqs |
|------------------------------------------------------------------+----------+------|
| Hacker News front page | 7.5kB | 5 |
| a Hacker News comment page | 31kB | 1 |
| “Life” on English Wikipedia | 570kB | 47 |
| Alexa.com's “top 500 global sites” page | 1.1MB | 69 |
| the “videos” page for HolaSoyGerman on YouTube (without a video) | 1.4MB | |
| my timeline on Twitter | 1.9MB | 80 |
| a NY Times editorial from March | 2.6MB | 179 |
| consumer electronics salesman's LinkedIn page | 3.1MB | 101 |
| my TL within a few minutes | 4.6MB | 466 |
| Amazon page for a deep-cycle lead-acid battery | 5.0MB | 193 |
Watching a high-resolution YouTube video can easily use 12 megabytes per minute.
So a median web page now is about 2MB of data in the initial load, and another several megs per minute. So it probably isn’t useful any more to keep the initial load below, say, ¼MB, and you can probably use 16 to 64 megs of data transfer to do the whole replication if the user doesn’t have to wait for all of that, but you probably don’t want to cost them 128 or 256 megs. I say this even though finding all of the above sites frustratingly slow, except for Hacker News, because apparently people use them anyway.
(Chris Zacharias’s famous “Page Weight Matters” tells about how he made YouTube much more useful in 2009 by reducing the Youtube watch page’s weight from 1.2MB to 98kB, enabling its use in areas in Siberia and Southeast Asia where 98kB still took two minutes to load. I think that probably in 2015 the number has gotten substantially higher.)