On 11/6/2010 9:08 AM, Dimitry Sibiryakov wrote:

> 06.11.2010 13:58, Geoff Worboys wrote:
>> Dimitry Sibiryakov wrote:
>>> Do you consider these points to be good? From this point
>>> of view, encryption on client side is an absolute winner
>>> as it solves all problems at once: network, memory, swap,
>>> hibernate, sort and storage are protected automatically. You
>>> just have to find encryption function which comply one rule:
>>> for every x1< x2, f(x1)< f(x2) must be also true.
>> I am not aware of any encryption that will satisfy your rule
>> and still provide good security.
> In this topic we seemed to agree that "good security" is impossible and some level of
> obscurity may be enough. In this case you can choose between, say, base64, uuencode, ROT
> or Caesar (if I remembered name for variable ROT right) algorithms.

No, this is not agreed. Good security is indeed possible. The question
is whether it is sufficiently desirable to implement.