Hi folks, I noticed that Cryptography-DES has no non-primitive version. Since the primitive is not in my current VM, I can't test it. Another issue is that interpreterProxy shows up as Undeclared when I load the DES stuff, which probably means that the primitive won't compile cleanly - but I have not looked further into this. Would you mind if I modify the DES code to fall back to the non-primitive implementation which I wrote for OpenPGP? I would have to adapt to the slightly different API anyway (my code used 8-byte Integers as keys and data blocks, while the Cryptography code uses 8-byte ByteArrays which make more sense).
BTW, I've made a first cut creating separate packages (Crpytography-Core, Cryptography-ASN1, ...) which seems to work reasonably well. Once I've successfully loaded all of them into a base image, I'll publish them to our repository. I don't know how MC manages prerequisites yet, so anyone loading these would have to take care of the load order by himself.
Cheers, Hans-Martin
In the Monticello Browser, you can right-click on a package and the pop-up menu has an 'add prerequisite' (or similarly named) entry.
On 10/23/05, Hans-Martin Mosner hmm@heeg.de wrote:
I don't know how MC manages prerequisites yet, so anyone loading these would have to take care of the load order by himself.
On Oct 23, 2005, at 5:11 AM, Hans-Martin Mosner wrote:
Hi folks, I noticed that Cryptography-DES has no non-primitive version. Since the primitive is not in my current VM, I can't test it.
Would you mind if I modify the DES code to fall back to the non-primitive implementation which I wrote for OpenPGP?
I was missing a non-primitive DES implementation, once upon a time. You can find the DES plugins for Windows and Linux here ( http://minnow.cc.gatech.edu/squeak/2410 ). No luck with the Mac plugin, which is why I can't squeak on my t-book. :`-(
Did you have to implement asn.1 for your OpenPGP implementation? I'll vote yes for moving that test case.
cheers, Robert
Wonderful Hans-Martin. Thank you for working on organizing the class structure.
I have the same issue with SHA256. There appears to be a primitive written for SHA1 for the expand and hash function, but since I have no experience in creating primitives, I didn't go that route.
I would think that we need two things to answer this question. First we need to quantify the benefits of the primitive. What kind of speed improvements do we actually get vs. the hiding of the code and making it harder to distribute, and making it more difficult to support. There would have to be major (orders of magnitude) speed improvements to make it worth it. Does anyone have primitive writing experience that would like to take this issue on?
In general I would say that every method that has a primitive call, (unless it would be totally impossible because of time limitations) should have a fall back method written. So yes I would say you should write the DES code.
Ron
-----Original Message----- From: cryptography-bounces@lists.squeakfoundation.org [mailto:cryptography-bounces@lists.squeakfoundation.org] On Behalf Of Hans-Martin Mosner Sent: Sunday, October 23, 2005 8:11 AM To: Cryptography Team Development List Subject: [Cryptography Team] DES status?
Hi folks, I noticed that Cryptography-DES has no non-primitive version. Since the primitive is not in my current VM, I can't test it. Another issue is that interpreterProxy shows up as Undeclared when I load the DES stuff, which probably means that the primitive won't compile cleanly - but I have not looked further into this. Would you mind if I modify the DES code to fall back to the non-primitive implementation which I wrote for OpenPGP? I would have to adapt to the slightly different API anyway (my code used 8-byte Integers as keys and data blocks, while the Cryptography code uses 8-byte ByteArrays which make more sense).
BTW, I've made a first cut creating separate packages (Crpytography-Core, Cryptography-ASN1, ...) which seems to work reasonably well. Once I've successfully loaded all of them into a base image, I'll publish them to our repository. I don't know how MC manages prerequisites yet, so anyone loading these would have to take care of the load order by himself.
Cheers, Hans-Martin _______________________________________________ Cryptography mailing list Cryptography@lists.squeakfoundation.org http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/cryptography
Ron Teitelbaum wrote:
... In general I would say that every method that has a primitive call, (unless it would be totally impossible because of time limitations) should have a fall back method written.
Good, that's exactly my POV, and it's very much in spirit with the original Smalltalk-80 primitive concept (they distinguished between essential and non-essential primitives, with Smalltalk fallback code for the latter ones).
So yes I would say you should write the DES code.
Ok, it will take some time since my representation is currently Integer-based and not ByteArray-based, but it should be done within this week. I will look at the other primitives, too.
Cheers, Hans-Martin
cryptography@lists.squeakfoundation.org
-
Cees De Groot -
Hans-Martin Mosner -
Robert Withers -
Ron Teitelbaum