Older blog entries for ajk (starting at number 26)

Got a visit from an angry amanuensis (is that the correct English term?) that we should leave the office we had been given ASAP (it was on loan from another department, and they needed the room for a foreign visitor). With some interdepartmentary diplomacy I was able to get us a new office, which is bigger and spacier than the previous one and is directly attached to the computer lab where we teach. I think it went quite well in the final analysis. The problem is, our new office space was reserved by a professor who has not used it in the last year, and we were given that space without asking the professor. Oh well, we need that office only for a month or so anyway. And I got my own work computer, on which I immediately installed Debian potato (as of first test cycle). The install system has really improved.

No time to code today, I am sorry to report. But yesterday I finally settled on two random number sources for Winters: /dev/random for the top-security stuff (keys etc) and a /dev/random-seeded MD5-based PRNG for the workhorse random numbers. And it works too, although I may have to rewrite the MD5 implementation (stolen from dpkg) for increased portability.

struct block decrypt(void * sc, struct block const bl)
{
        struct block rv = { 0, NULL };
        return rv;
}

That's what caused the zero-size decryption result. I feel so damn dumb now.

Wrote some more of Winters networking code. For some strange reason the receiver decryption result for the packets have size zero. I need to investigate that.

My copy of Schneier came on Tuesday, have been reading it enthusiastically.

I am offering some Debian packages for adoption by registered Debian developers or sponsored prospective developers.

Today started both summer courses: the programming course (where I work as a teaching assistant) and the topology course (where i sit and study). The next one and half months is going to be busy time for me.

Made an offline copy of the potato i386 archive for my father, and installed a basic potato system for him. This took all my free time today, so I did not have any time to code or spec.

I've been specing and implementing Winters network protocol. It's going slowly. I managed to write my own Blowfish implementation which seems to be correct (passes a simple encryption/decryption test where I know the correct ciphertext). I noted, however, that the online description of Blowfish is ambiguous in many occasions, most notably it omits information on endianness issues on many occasions. Apparently my guess "big-endian über alles" was correct. My Blowfish implementation is not one of the fastest, but it is the only one that I know that does not require separate source configuration (either manually or with Autoconf) and is written in C.

Did you know that GCC means "Graphic Character Combination"? I didn't, until I ran across the ECMA-48 standard.

I should've known. I presented the session management part of the Winters protocol to a security newsgroup. It was pretty much shot down. They even asked why I was designing my own protocol instead of using SSL or SSH or some other well-established protocol. The answer is that they don't take advantage of one unique aspect of the application I'm designing the protocol for: all legitimate sessions are necessarily derived from a root session. My protocol thus can do without all long-term keys, symmetric or asymmetric. Oh well, back to the drawing board (and I'm getting a copy of Schneier).

I decided to rebuild my home system, which naturally runs Debian, from scratch to do away with all its cruft. Of course I forgot that I need the drivers diskette even in the NFS install, and of course I noticed that after wiping out my previous install. I had no access to the network anymore. I had several past versions of Debian on CD's but alas, my CD drive is broken. I even had some old boot disks for Debian 1.3, Debian 0.93r6, and Slackware 3.0. Unfortunately, none of them would build my system into a shape where it could reach out to the net to grab the stuff I needed. Ended up calling a nearby friend and using his machine for creating that much needed drivers diskette.

Today my network connection suddenly dropped. Turned out my network card was loose and had disconnected.

I've been drafting the Winters network protocol. I found uses to such 32-bit quantities as 4037931181 ("emergency hangup") and 4026658817 ("permission denied") [thanks dhd!]. The protocol is simple and expresses what it needs to express. I realized I can set up a separate auth server to provide new sessions for arbitrary clients if necessary, which simplifies the problem. You see, the protocol creates new sessions by negotiating their cipher and encryption key on an existing session; this avoids me having to deal with public-key cryptography in the protocol itself.

Anybody see any obvious security problems with assuming that if the client can send a known plaintext (published as part of the protocol specification) encrypted with the correct key (which is prenegotiated using channels we assume to be secure) then the client can be considered authenticated?

The next step is to implement the protocol. This should not take more than two days (famous last words!).

Advertised Advogato to Tuomas Lukka. Seems he likes it.

In the spirit of the recent article on project naming, I decided to rename TeWiS into Winters. Sounds much better that way.

schoen, you can find a list of recommended compiler technology books in the comp.compilers FAQ. The book is the dragon book, Compilers: Principles, Techniques, and Tools by Aho, Sethi and Ullman (Addison-Wesley, 1986). I've also found Wirth's books on the topic instructional (he has a book that, IIRC, builds a complete Pascal compiler as it goes, and I believe his Data Structures + Algorithms has a section on recursive descent). Stroustrup builds an expression calculator using recursive descent in his C++ bible (as do many other authors in various books using various languages).

It is pretty straightforward to write a recursive descent parser based on a BNF grammar by hand: make every nonterminal a procedure that parses that nonterminal, and have it recursively invoke other procedures whenever the BNF grammar refers to other nonterminals. Parse terminal symbols any way you see fit (a separate lexer that tokenizes the input is easy to write ad hoc or by using regular expressions). See any language/automata theory book or compiler technology book for more info.

In other news

I have taken over the TeXmalli document from liw. It is a sample LaTeX document written in Finnish.

I have been so involved in real life since Friday that I haven't had a chance to post any diaries. Most of that was because I was moving (new real-life contact information on my home page) and thus have acquired a permanent net connection, so I hope I am excused :-)

For the same reason, I have not been able to do any development work in the last few days.