Re: [Python-Dev] Support for Encrypted Zip as python scripts
I like the idea, but... Here is a quick list of things to think about and if some of this has already been mentioned, sorry. Speed: Encryption speed has been mentioned. For short scripts this may not be a problem, although algorithms implemented in C would be faster. Strength: Passwords are [very] weak, especially if of the 6-10 alphanumeric variety. True secret keys where all bit combinations are used is stronger. Entering passwords has been mentioned but I believe only passwords were assumed. It is better to not provide any encryption than to lure novices into believing they are secure when they are not. Algorithms: Be sure to choose good ones and allow for changing later. Key distribution: How to distribute secret keys beyond a small group of friends is problematic. In short it doesn't scale. Looking to public-private key pairs can be equally problematic. This can get you into encryption certs, but *how* you use them correctly differs from signing certs. More on this later if you want. ZIP: Look beyond just zip files. A scheme that works for any/all files in the distribution, not just ZIPs, would be better. (IIRC there have been problems with encrypted zips, but that was years ago. Those issues may have been fixed.) Short version: Doing this right is hard. Simply supporting a password based ZIP file is, in my opinion, not real protection. Gotta go. Later. Larry
On Mon, Aug 24, 2009 at 11:52 AM, Bugbee, Larry
I like the idea, but...
For what use case? -- --Guido van Rossum (home page: http://www.python.org/~guido/)
I like the idea, but... For what use case?
I don't have a specific case in mind. In general, however, it would be nice to be able to protect intellectual property, but without addressing the problem from a holistic view, there is little protection afforded and perhaps a lot of unrewarded work. And I forgot one, Distribution of crypto across certain international borders. Export/import laws by itself can be a showstopper.
Bugbee, Larry wrote:
I like the idea, but... For what use case?
I don't have a specific case in mind. In general, however, it would be nice to be able to protect intellectual property
This I'm also unclear about. How does it protect intellectual property? Won't the person running the zipfile have to enter the password? Whom would you protect the IP from? Regards, Martin
I like the idea, but... For what use case? I don't have a specific case in mind. In general, however, it would be nice to be able to protect intellectual property This I'm also unclear about. How does it protect intellectual property? Won't the person running the zipfile have to enter the password? Whom would you protect the IP from?
I agree the IP will have to be exposed at some point to be useful, but let's not overlook other things that could be in play like PIA agreements and the like. Also, something stronger than a password will be needed to be secure, and secret key distribution does not scale. There is a lot more to consider and we are only scratching the surface. Confidentiality in-the-large will take far more than encrypted ZIP files. Please know that I am not pushing for the encryption of ZIP files and this thread is going down a path I did not intend, or desire pursuing. My original post was intended to increase the awareness in those thinking encrypted ZIP files will 1) be easy, 2) afford the protection they desire, and 3) not lead others into a sense of false security. Encryption sounds good, but doing it right can be a landmine. A quick fix to support ZIP files will likely create more problems than it will solve. I still say it would be *nice* if there was some way to protect IP. I have no expectations that it will be easy, and least of all, solved by encrypted ZIP files and a simple patch to Python. ...but that should not diminish the desire. Larry
Bugbee, Larry writes:
My original post was intended to increase the awareness in those thinking encrypted ZIP files will 1) be easy, 2) afford the protection they desire, and 3) not lead others into a sense of false security.
All good points, but note that (even without the DMCA) at least in the U.S. copyright law provides for *criminal* penalties for willful infringement. If you needed to type in a password to copy, you can't argue "I didn't know this was private property", just as crossing a fence is stronger evidence of criminal trespass than ignoring "Posted" signs. For patents, the password prompt could notice of patent protection, with similar effect of strengthening penalties for infringing. So even weak encryption strengthens the available legal protection. That is not sufficient reason to consider putting encrypted zips or anything similar into the stdlib. It's relevant to users' decisions should such features become available, that's all.
I still say it would be *nice* if there was some way to protect IP.
-1. Intellectual assets *can* give benefits with zero further costs of production and almost negligible costs of distribution. But IP, like any other property that requires a temporary transfer of possession to give economic benefit (eg, rental cars), is going to involve substantial transaction cost for consumers (search for the product, license negotiation[1]), as well as the usual excess burden of monopoly. The current state where only legal protection is feasible is arguably a good compromise. Since it involves substantial costs of enforcement borne by the rightsholder, it's only going to be invoked where the total social benefit (net consumer value plus vendor profit) is large enough to swamp the small transaction costs. I think that Python should spend zero effort on implementing technical means of IP protection. Any side effects of privacy protection devices should be more than enough to serve. Regards, Footnotes: [1] Not necessarily bargaining, but also including studying the terms of take it or leave it offers, etc.
On Mon, Aug 24, 2009 at 2:01 PM, Bugbee, Larry
I don't have a specific case in mind. In general, however, it would be nice to be able to protect intellectual property, but without addressing the problem from a holistic view, there is little protection afforded and perhaps a lot of unrewarded work.
I would think just distributing pyc files would achieve that goal
participants (5)
-
"Martin v. Löwis" -
Bugbee, Larry -
Dj Gilcrease -
Guido van Rossum -
Stephen J. Turnbull