Refreshing incremental backup (using digital signatures)

I originally felt that, do incremental backup, is to calculate each segment of this file (such as 20M size) of the digital signature every time, if each segment of the digital signature and the previous calculation is exactly the same, then you do not have to upload the paragraph, just upload the modified segment. Of course, after being modified, the subsequent segments will not be aligned, so you will have to calculate the digital signature (high computational strength) of the data that starts at 20M from it along each byte, until the same digital signature is found, so that you can save uploading the paragraph. After the calculation found that the two-segment inconsistency after the use of the flow technology positioning began to rewrite, has been rewritten to two large files and the same paragraph so far. If there are different segments after the same segment, repeat the previous procedure. This approach doesn't look difficult, but if you do, the whole process can be tedious.

-------------------------------------------------------------------------------------------------

Now I have heard another saying that this approach takes full advantage of the results of mathematics, not the simple logic that I use:
http://librsync.sourcefrog.net/
This means that after the first upload, the file will never change, and each time the local file is modified, the difference between it and the first file can be calculated. This way, no matter how many times the file is modified, it will always be a raw file and an incremental file, and it will be able to generate new files together. When uploading, always upload only that delta file.
The principle looks very simple, but the difference is how it is calculated, this is very important. Because not only do you want to record new data, but also to record the location of the new data relative to the original file and what to replace. Regardless of how the original file is changed, the new file is calculated directly relative to the first original file's digital signature (instead of the third new file's digital signature relative to the second new file). If the original file is constantly changing, the information to be recorded is much more slowly, and the efficiency of uploading is lower. But it's still a simple and easy-to-use incremental backup solution.

If you want to dig the potential of incremental backup, then after each upload, send an instruction, the last upload of the original file and the delta file merge, recalculate its digital signature, then the next incremental upload can be based on the file above once, rather than the first upload of the file as a long time ago.

Refreshing incremental backup (using digital signatures)