Hacker News new | comments | show | ask | jobs | submit login

One reason for extra large keys is for headroom in case of partial breakage.

Usually when a cipher is broken, they don't break it fully, but rather reduce the keyspace. By making your starting keyspace even larger you make even a broken cypher secure (up to a point).

You are correct but its also extremely educational to graph the decrease over time, for various cryptosystems both real/pro and "roll your own".

The difference between a professionally designed algo and a "roll your own" algo is the roll your own is usually extremely brittle and shatters to nothing in one discovery vs a professionally designed algo usually has multiple generations of grad students chip away maybe an average of 10 bits per decade as newer forms of analysis are slowly invented and/or declassified.

If you roll your own, its at least several orders of magnitude more likely to break than a pro algo, and when it does, unlike the pro algo, it'll almost certainly snap all the way at once, so using a large keyspace just means more of an epic fail and wasted processor power. On the other hand if you use a real algo, then a loss of 20 bits per decade or whatever average really does help, so crank the keyspace to the max. So I'd have to disagree with you there, if you're using a broken (homemade?) algo don't bother with more than 8 bits or so, any more is just wasting processor time, homemade is only useful for casual obfuscation. If you're using a real algo, then, and only then, larger keyspace actually will save you over the long run...

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact