I did set up ArsDigita so that every employee got 5 weeks of vacation annually (enough for a 3-week break in the summer plus two one-week breaks). Maybe I should add that to the article.
The biggest ACS communities of which I'm aware are about 600,000 users (see away.com for example).
About 1/3rd of ArsDigita customers started as raw startups (www.infirmation.com, away.com, uslaw.com (recently linked as AltaVista's exclusive Law channel)). We've watched them grow. I wish I could tell you that they've all started to turn a profit because ACS is so great. But as far as I know they are all more concerned about building market value, company size, etc. Many of them seem to be getting acquired for big bucks, raising additional stupefying sums of venture capital, etc. So they are achieving their business objectives but I doubt that very many of our startup customers have to pay income tax.
Anyway, ACS per se can scale quite nicely if you budget for a big RDBMS server and a rack of small Web servers and a load balancing switch (about $300,000 all together). But the users might not be happy because there will be too many comments on each page, etc. It is up to the publisher to use the user groups features intelligent to split up the community a bit.
I'd love to hear from you privately, blackdefiance. I doesn't bother me too much that you thought aD was arrogant (my brother and Cesar are both Harvard grads so they probably can't help it). But I'm very disturbed by the idea that our tech staff couldn't solve a problem. Jin Choi, for example, added Web-based email to the ACS and it took him only a couple of weeks (he's just finishing up an ArsDigita Systems Journal article on the subject so you can find out how). Bottom line is that we've never had a tech problem that we couldn't solve so I'd like to know how this impression was communicated (exception: some sites like my personal ones have been resource-constrained so we haven't been willing to throw in enough bodies to keep Oracle Intermedia Text working, for example).
So give me a call at 617-386-4112 or email philg@mit.edu and let me know what we did wrong. I promise to send you one of my prints and an aD fleece (they are very nice!).
Good question! I actually have some (2 or 3) powerful VA Linux boxes spinning in my office at MIT. Oracle 8.1.6 came out on Solaris in December. I'm still waiting for the Linux version (current theory is next week on technet.oracle.com). Oracle 8.1.5 is basically impossible to install on Linux and there are a lot of nice new features in 8.1.6. But these are fast machines, really. They are dual P700s. Should be faster than the doom-wankers' old machine (creaking 4-CPU 167 MHz SPARC/Solaris) but nothing to impress the high-volume crowd (but pretty powerful for 2U of rack space, I hope!)
If so, I'm a big believer. That's what we had on the old MIT Lisp Machine back in the late 1970s and it was much more productive than programming with raw tools. The thing that the Lisp Machine lacked was a database for persistent storage and hence it really isn't practical to use that old environment for developing Web apps.
So anyway, I love great programming tools and wish that we had better ones for the work that I do. But I don't like to get religious about it. A really good programmer in the end will be able to debug a system with print statements.
Yes, broody, all of our undergrads actually do learn enough Tcl and AOLserver in a day. They spend about 15 hours doing http://photo.net/teaching/psets/ps1/ps1.adp and they end up having to learn Tcl and AOLserver programming as a subtask (we never explicitly teach it them; they are expected to pick it up).
The whole point of Tcl is that you're not supposed to have to be a programmer to use it! Read Ousterhout's original paper on why scripting languages are a good idea.
This isn't a theoretical point. We've taught software engineering for Web apps to several hundred people now. We've used AOLserver/Tcl precisely because we don't have to spend any class time teaching it. We could use a Java version of our toolkit for future courses but I'm not sanguine about the prospects of introducing Java as a two-hour subtask during pset 1.
Anyway, we do have marketing shills at ArsDigita now, but I feel like the Slashdot community deserves a response from our tech staff (of whom I'm still titular head).
Bottom line as far as I can tell is that there are a lot of folks on Slashdot who want to believe that they are smart even though they received poor SAT scores and did not graduate from college. They think that ArsDigita U is passing judgement on people who didn't graduate from college or got trashed on the SATs. But it isn't so! We're simply trying to copy Harvard and MIT as much as practical and thereby reduce our risk of failure. I didn't even graduate from high school so I'm hardly in a position to call people losers because they don't have impeccable credentials.
Doh! Those of you who noticed how badly hosed photo.net was for an hour today might be amused to know that I slashdotted my own site. Most of the stuff on photo.net is articles illustrated with photos. So if someone grabs an article they get enough text to read for 10 minutes plus 20 photos as illustration (21 hits total). What I did in my Slashdot interview was link to some PhotoCD index pages. These have no text on them but reference 108 inline JPEGs each. So every Slashdot user who followed one of those links was generating 109 hits immediately on the photo.net server, reaching peak requests of many hundreds of hits/second (too much for a 180 MHz old HP-UX machine).
I feel like an idiot. Generally I never publicly link these pages but just use them for my convenience in developing new content.
We should be more clear on the Web page... the 1400 SAT scores are just a minimum to make sure that we get students vaguely like those at MIT or Harvard. After we get the minimum, we look at the essay and then we talk to the person (phone interview or maybe in-person as well).
Okay, okay. If you want one Lisp app that is useful... Yahoo! Store. It does the job for its customers and it was sold by its developers for $50 million or so (a few years ago when $50 million was still considered real money).
On the Emacs front, it just happens to be the text editor that everyone I know uses so I employed the term "Emacs" for concreteness. Obviously Emacs per se doesn't help you understand a data model (though I do like to run SQL*Plus in an Emacs shell because it is easy to cut and paste queries).
Anyway, I'm not a Lisp programmer anymore! I write Java, PL/SQL, SQL, (gasp) Tcl, and sometimes even Perl!
I like the Zope guys too. But if you can figure out how to serve 1 million hits/day PLUS the slashdot overflow on a computer with 180 MHz processors (that is also doing 20 other sites at the same time and running Oracle), please tell me how (remember that each page runs several RDBMS queries).
This does suck. I guess it shows that a single computer with 180 MHz CPUs can't handle the RDBMS plus the Web server plus 20 other sites... We were planning to upgrade soon anyway.
It is dangerous to be a harsh reviewer of student projects in the age of universal Web connectivity (and anonymity)! (Assuming this person really is one of our students, which I'm not sure that he is because we told our students this term about the Java version of ACS and we never celebrate Tcl except as an example of how the Lisp community's avoidance of the users' problems led to oblivion)
As for the course teaching theory, we do teach the students about transactions and concurrency and abstraction but what the poster forgets is that this is a LAB COURSE. It is supposed to be about problem sets and projects and not theory. We have other courses at MIT that teach theory, some of which I've taught.
I guess I'm pretty sure now that this person isn't a real MIT student because he says that the course teaches no "practical knowledge". The students come in not knowing what an RDBMS is. After 13 weeks, they are proficient in SQL and know some of the intracacies of Oracle.
Bottom line is that people come in not knowing HTTP, HTML, or SQL, and go out having built systems like www.arfdigita.org (check it out). They must have learned quite a bit because they were helpless to build any kind of Web service (db-backed or otherwise) when they came in.
I'm already at the ArsDigita London office (got here on Friday night) and I must say that English ISDN lines feel slower than my old 14.4 modem in the US. The only reason British Telecom can do this to you guys is that the government has taken away your guns. We may have mass killings in the US but at least the firepower seems to have cowed utility monopolies into a semblance of service.
Our Tokyo office right now is... Henry Minsky (yes, he is the son of Marvin Minsky). You can reach him at hqm@arsdigita.com
We're trying to grow as fast as possible in Japan so I'm sure that Henry would be delighted to hear from you. And it is certainly staffed locally because we don't have a surfeit of Japanese speakers around Cambridge...
It certainly does not make sense to require folks to have a programming background. MIT does not require a programming background for admission to the EECS undergrad program.
As for "struggle with the problems" our students will have ample time for that. Only 2 or 3 hours per day (out of 12) are spent in lecture. The rest if problem sets.
When will our students have time to "hack on large projects"? The rest of their frigging lives! We want to do the best that we can with one year and then send our graduates on their way. We don't expect them to become Richard Stallman-grade programmers during one year with us or anyone else.
Finally, who ever said that we were teaching programming? We're teaching computer science! Some of our students might choose to go on to grad school at CMU and bury themselves in complexity theory or parallel algorithms or whatever. (We do have a couple of courses in software engineering but those are only two out of 11).
We're sort of making up the rules as we go along (this is the first year, remember). We probably won't formally close off applications until August. We will admit some obviously qualified and committed people "early decision" to give them time to clean up their personal affairs. We've only had about 50 applications so far (last Friday's san Jose Mercury News article was the first time anyone had heard of us). Of these, it looks like about 5 are obvious admits.
I had a B+ average at MIT undergrad (math IS hard, as Barbie noted) so I'm not really in a position to call someone without straight-As a pinhead. That said, someone who got Cs and below in a bunch of courses would make us worry a bit.
We look closely at the essay and also talk to the applicants. We're very interested that the student have some goal in life about which he or she is passionate. Even at MIT, it can be depressing to teach a bright student who is simply going through the motions.
Your suggestion that we try to do more than a post-bacc is a good one. But keep in mind that we don't have much money (only $1 million/year). It is a lot easier to teach a Harvard grad than a raw high school kid. We're going to do the easy stuff first, try to do it well, and then expand as we get more resources.
If you look at Michael Saylor, who started his school with $100 million, you see the difference that more money makes. He is trying to do a whole liberal arts curriculum. He is trying to innovate in the method of instruction (online instead of face-to-face). I predict that he will have a tougher time than we will, even though he isn't constrained by money. When looking for teachers across a broad range of subjects, he will have a tough time just figuring out whom to call. Whereas I'm only one degree of separation from any qualified CS teacher in the world.
If we were to start giving bachelor's degrees we'd have to ask ourselves in what way we were better than Harvard or MIT. Those schools have, respectively, about 350 and 150 years of experience doing what they do. We would be tuition-free but so will they in the long run (I think). The guy below who said that I was an egomanaic is mostly right. I don't like to be involved in something unless it can be the world's best. I know how to make the world's best post-bacc CS program. But I don't know how to build a great college that will do all the things that need to be done for 18-22 year olds.
Maybe more than a couple will join ArsDigita but really we don't want people unless they are hardcore about http://arsdigita.com/mission/. ArsDigita is just not a very fun place to work for people who want to have an easy life, weekends with the kids, etc. To be a great ArsDigita one must (1) build great software, (2) come up with innovations, (3) teach others how to practice those innovations via short papers, longer books, and face-to-face courses.
A guy with an MD or a mother returning to the workforce after a bunch of years home with some kids might not want to subscribe to our brand of fanaticism. And they won't really be immersed in our culture as you suggest. More than half the faculty are not employees of ArsDigita (we have our share of CS PhDs but we've got work to do so we're also hiring CS PhDs whose first love is teaching).
We attack the recruiting problem with little sports cars and other, more immediate, incentives. But ultimately the best thing that we have to offer people is the guarantee that they'll be working with other smart folks (our Pasadena office was started by four Caltech PhDs) and that they'll be free to build great things without meddling from clueless managers or designers.
I think that the reason colleges do 4 courses in parallel instead of intensively is because it leads to maximum admin convenience given (1) full-time faculty, and (2) students in different majors or on different schedules. Note that MIT was built during the Industrial Revolution and works sort of like a facotry. The professors are like machines. They are nailed to their classrooms. The students are like the widgets. They move from machine to machine on conveyors (corridors). Compare to Oxford University, started during the time of guilds and apprenticeships. They have tutors.
1300 on the SATs might be good enough if you took them before the rescaling. Your HS career is certainly better than mine! (see photo.net/samantha/ for an account of my high school achievements). Anyway, we don't stop people from taking the SATs again and crushing them like a bug.
Why do we focus on SATs? MIT has a huge admissions bureaucracy that does a great job but costs more money to run than our whole little school. So we needed some criteria that would yield us a high-quality student body but that wouldn't take up a lot of time to administer.
I thought I was done with my PhD Thesis Defense...
on
ArsDigita University
·
· Score: 2
... and yet here is one of the faculty members who tortured me in grad school, coming into slashdot anonymously. It hurts.
If I were intelligent or level-headed, I would have gone to medical school like my brother Harry.
But rest assured, ArsDigita University students won't have to suffer with me all year. I'm only teaching one course (though I might be a TA for some of the others).
(Note to the rest of you: before you start a little online community like photo.net, be aware that you might get "a few dozen" unsolicited emails over the years from someone like this anonymous coward. It is one of the great joys of being a non-commercial Web publisher.)
Lisp hackers were too busy arguing with Perl hacke
on
ArsDigita University
·
· Score: 3
Oh yes, the most interesting question you raise is actually answered in a cut-and-pasted excerpt from my Tcl for Web Nerds intro. Why weren't there any good Lisp-based Web servers in 1995 (or now)? Lisp programmers were too busy congratulating themselves for being smarter than C programmers. So the C programmers sat down and figured out what the real problems of publishers and end-users were and came up with practical stuff like AOLserver in 1994, Apache/mod_perl (1998 or 1999 by the time all the db connection pooling kinks were worked out?), etc.
Engineering excellence doesn't mean having a fancier system; it means having a system that solves the users' problem better faster and cheaper.
Am I bitter? No worse than Medea...
(I'm also confused by the person who claims that pgreesnpun isn't me. I am me, dammit!)
Most of the hackers at ArsDigita (including moldy old me) have built Web presentation layers in Perl, Tcl, and Java. We agree with you that Java isn't an improvement over Tcl for merging templates with db data. On the other hand, since the presentation language isn't core to our toolkit, why not make our data models and abstractions available to those who've got big Java libraries (and, okay, the ArsDigita investor in me says "to those who have big budgets")?
Why didn't AD select a Web server that uses Lisp? There weren't any back in 1995 that could talk to an RDBMS. We wanted to build applications to solve real users' problems, not get mired in a tools debate. So we built apps instead of tools. SAP was written in COBOL initially. I'm sure that they didn't like COBOL but they wanted to solve the ERP instead of the language problem so they did (and now SAP is the world's #2 market cap software company, after Microsoft).
Why was I snide? I'm sick of hearing arguments about languages and development tools. It is what gives programmers a bad reputation for being losers.
A few MIT things? The operating system. Time-sharing. RSA encryption. TCP/IP and a lot of earlier network stuff. Computer algebra (and the Lisp-based Macsyma, which is still the most useful computer algebra system (though I admit that Maple is easier to install and use)). As for Lisp, a lot of language bigotry caused Lisp programmers to waste time arguing and reimplementing Lisp instead of building apps. But some good apps were built, esp. in the areas of electronic design and automated mechanical design. A lot of systems used Lisp as an extension language, e.g., the Interleaf publishing system and the AutoCAD drafting system. And a lot of the good ideas from Lisp made it into Java.
But the bottom line is that tools per se won't make much difference. At MIT, we teach our students to generate entire Web sites (including all the scripts) from machine-readable specs. Instead of arguing over tools and maybe getting a 3X productivity improvement (e.g., from using Apache/Perl instead of Netscape Enterprise Server/Java), they are writing programs that write programs. They'll get a 1000X productivity improvement over even the best Perl coder even if their ultimate target language is EDSAC machine code.
The future of Web development is high-level specification and automatic generation of sites. This is what we try to teach students, where we do at ArsDigita when appropriate for a service, and where we are pushing toolkit development.
We'd love to be accredited but it is tough when ours is the only program of its kind. Accreditation doesn't make much sense without a basket of schools teaching the same sort of curriculum.
We don't really want to teach people without a bachelor's degree. If someone doesn't have a bachelor's, he or she can simply apply to MIT and get a regular SB in EECS!
Fear not (you can always use Macsyma and Maple)
on
ArsDigita University
·
· Score: 2
Fear not! As noted below, we're only going to do the first course in calculus from MIT (18.01, which is a few weeks of differential calculus and then the rest of the term on integral calculus). The last thing a CS nerd needs is Div, Grad, or Curl. Nor do they need to know how to use a Frobenius to approximate a solution to a differential equation!
And if the students don't learn calculus, they can always become lamers like me and use Maple or Macsyma:-)
Slashdot Mirror
I did set up ArsDigita so that every employee got 5 weeks of vacation annually (enough for a 3-week break in the summer plus two one-week breaks). Maybe I should add that to the article.
The biggest ACS communities of which I'm aware are about 600,000 users (see away.com for example).
About 1/3rd of ArsDigita customers started as raw startups (www.infirmation.com, away.com, uslaw.com (recently linked as AltaVista's exclusive Law channel)). We've watched them grow. I wish I could tell you that they've all started to turn a profit because ACS is so great. But as far as I know they are all more concerned about building market value, company size, etc. Many of them seem to be getting acquired for big bucks, raising additional stupefying sums of venture capital, etc. So they are achieving their business objectives but I doubt that very many of our startup customers have to pay income tax.
Anyway, ACS per se can scale quite nicely if you budget for a big RDBMS server and a rack of small Web servers and a load balancing switch (about $300,000 all together). But the users might not be happy because there will be too many comments on each page, etc. It is up to the publisher to use the user groups features intelligent to split up the community a bit.
I'd love to hear from you privately, blackdefiance. I doesn't bother me too much that you thought aD was arrogant (my brother and Cesar are both Harvard grads so they probably can't help it). But I'm very disturbed by the idea that our tech staff couldn't solve a problem. Jin Choi, for example, added Web-based email to the ACS and it took him only a couple of weeks (he's just finishing up an ArsDigita Systems Journal article on the subject so you can find out how). Bottom line is that we've never had a tech problem that we couldn't solve so I'd like to know how this impression was communicated (exception: some sites like my personal ones have been resource-constrained so we haven't been willing to throw in enough bodies to keep Oracle Intermedia Text working, for example).
So give me a call at 617-386-4112 or email philg@mit.edu and let me know what we did wrong. I promise to send you one of my prints and an aD fleece (they are very nice!).
Good question! I actually have some (2 or 3) powerful VA Linux boxes spinning in my office at MIT. Oracle 8.1.6 came out on Solaris in December. I'm still waiting for the Linux version (current theory is next week on technet.oracle.com). Oracle 8.1.5 is basically impossible to install on Linux and there are a lot of nice new features in 8.1.6. But these are fast machines, really. They are dual P700s. Should be faster than the doom-wankers' old machine (creaking 4-CPU 167 MHz SPARC/Solaris) but nothing to impress the high-volume crowd (but pretty powerful for 2U of rack space, I hope!)
IDE = integrated development environment?
If so, I'm a big believer. That's what we had on the old MIT Lisp Machine back in the late 1970s and it was much more productive than programming with raw tools. The thing that the Lisp Machine lacked was a database for persistent storage and hence it really isn't practical to use that old environment for developing Web apps.
So anyway, I love great programming tools and wish that we had better ones for the work that I do. But I don't like to get religious about it. A really good programmer in the end will be able to debug a system with print statements.
Yes, broody, all of our undergrads actually do learn enough Tcl and AOLserver in a day. They spend about 15 hours doing http://photo.net/teaching/psets/ps1/ps1.adp
and they end up having to learn Tcl and AOLserver programming as a subtask (we never explicitly teach it them; they are expected to pick it up).
The whole point of Tcl is that you're not supposed to have to be a programmer to use it! Read Ousterhout's original paper on why scripting languages are a good idea.
This isn't a theoretical point. We've taught software engineering for Web apps to several hundred people now. We've used AOLserver/Tcl precisely because we don't have to spend any class time teaching it. We could use a Java version of our toolkit for future courses but I'm not sanguine about the prospects of introducing Java as a two-hour subtask during pset 1.
Anyway, we do have marketing shills at ArsDigita now, but I feel like the Slashdot community deserves a response from our tech staff (of whom I'm still titular head).
Bottom line as far as I can tell is that there are a lot of folks on Slashdot who want to believe that they are smart even though they received poor SAT scores and did not graduate from college. They think that ArsDigita U is passing judgement on people who didn't graduate from college or got trashed on the SATs. But it isn't so! We're simply trying to copy Harvard and MIT as much as practical and thereby reduce our risk of failure. I didn't even graduate from high school so I'm hardly in a position to call people losers because they don't have impeccable credentials.
Doh! Those of you who noticed how badly hosed photo.net was for an hour today might be amused to know that I slashdotted my own site. Most of the stuff on photo.net is articles illustrated with photos. So if someone grabs an article they get enough text to read for 10 minutes plus 20 photos as illustration (21 hits total). What I did in my Slashdot interview was link to some PhotoCD index pages. These have no text on them but reference 108 inline JPEGs each. So every Slashdot user who followed one of those links was generating 109 hits immediately on the photo.net server, reaching peak requests of many hundreds of hits/second (too much for a 180 MHz old HP-UX machine).
I feel like an idiot. Generally I never publicly link these pages but just use them for my convenience in developing new content.
Doh!!!!!!!!
We should be more clear on the Web page... the 1400 SAT scores are just a minimum to make sure that we get students vaguely like those at MIT or Harvard. After we get the minimum, we look at the essay and then we talk to the person (phone interview or maybe in-person as well).
Okay, okay. If you want one Lisp app that is useful... Yahoo! Store. It does the job for its customers and it was sold by its developers for $50 million or so (a few years ago when $50 million was still considered real money).
On the Emacs front, it just happens to be the text editor that everyone I know uses so I employed the term "Emacs" for concreteness. Obviously Emacs per se doesn't help you understand a data model (though I do like to run SQL*Plus in an Emacs shell because it is easy to cut and paste queries).
Anyway, I'm not a Lisp programmer anymore! I write Java, PL/SQL, SQL, (gasp) Tcl, and sometimes even Perl!
I like the Zope guys too. But if you can figure out how to serve 1 million hits/day PLUS the slashdot overflow on a computer with 180 MHz processors (that is also doing 20 other sites at the same time and running Oracle), please tell me how (remember that each page runs several RDBMS queries).
This does suck. I guess it shows that a single computer with 180 MHz CPUs can't handle the RDBMS plus the Web server plus 20 other sites... We were planning to upgrade soon anyway.
It is dangerous to be a harsh reviewer of student projects in the age of universal Web connectivity (and anonymity)! (Assuming this person really is one of our students, which I'm not sure that he is because we told our students this term about the Java version of ACS and we never celebrate Tcl except as an example of how the Lisp community's avoidance of the users' problems led to oblivion)
As for the course teaching theory, we do teach the students about transactions and concurrency and abstraction but what the poster forgets is that this is a LAB COURSE. It is supposed to be about problem sets and projects and not theory. We have other courses at MIT that teach theory, some of which I've taught.
I guess I'm pretty sure now that this person isn't a real MIT student because he says that the course teaches no "practical knowledge". The students come in not knowing what an RDBMS is. After 13 weeks, they are proficient in SQL and know some of the intracacies of Oracle.
Bottom line is that people come in not knowing HTTP, HTML, or SQL, and go out having built systems like www.arfdigita.org (check it out). They must have learned quite a bit because they were helpless to build any kind of Web service (db-backed or otherwise) when they came in.
I'm already at the ArsDigita London office (got here on Friday night) and I must say that English ISDN lines feel slower than my old 14.4 modem in the US. The only reason British Telecom can do this to you guys is that the government has taken away your guns. We may have mass killings in the US but at least the firepower seems to have cowed utility monopolies into a semblance of service.
Our Tokyo office right now is ... Henry Minsky (yes, he is the son of Marvin Minsky). You can reach him at hqm@arsdigita.com
We're trying to grow as fast as possible in Japan so I'm sure that Henry would be delighted to hear from you. And it is certainly staffed locally because we don't have a surfeit of Japanese speakers around Cambridge...
It certainly does not make sense to require folks to have a programming background. MIT does not require a programming background for admission to the EECS undergrad program.
As for "struggle with the problems" our students will have ample time for that. Only 2 or 3 hours per day (out of 12) are spent in lecture. The rest if problem sets.
When will our students have time to "hack on large projects"? The rest of their frigging lives! We want to do the best that we can with one year and then send our graduates on their way. We don't expect them to become Richard Stallman-grade programmers during one year with us or anyone else.
Finally, who ever said that we were teaching programming? We're teaching computer science! Some of our students might choose to go on to grad school at CMU and bury themselves in complexity theory or parallel algorithms or whatever. (We do have a couple of courses in software engineering but those are only two out of 11).
I'm in our London office now, as a matter of fact. We're expanding Munich and Tokyo. We'll be opening Sydney and maybe Paris Real Soon Now (TM).
We're sort of making up the rules as we go along (this is the first year, remember). We probably won't formally close off applications until August. We will admit some obviously qualified and committed people "early decision" to give them time to clean up their personal affairs. We've only had about 50 applications so far (last Friday's san Jose Mercury News article was the first time anyone had heard of us). Of these, it looks like about 5 are obvious admits.
I had a B+ average at MIT undergrad (math IS hard, as Barbie noted) so I'm not really in a position to call someone without straight-As a pinhead. That said, someone who got Cs and below in a bunch of courses would make us worry a bit.
We look closely at the essay and also talk to the applicants. We're very interested that the student have some goal in life about which he or she is passionate. Even at MIT, it can be depressing to teach a bright student who is simply going through the motions.
Your suggestion that we try to do more than a post-bacc is a good one. But keep in mind that we don't have much money (only $1 million/year). It is a lot easier to teach a Harvard grad than a raw high school kid. We're going to do the easy stuff first, try to do it well, and then expand as we get more resources.
If you look at Michael Saylor, who started his school with $100 million, you see the difference that more money makes. He is trying to do a whole liberal arts curriculum. He is trying to innovate in the method of instruction (online instead of face-to-face). I predict that he will have a tougher time than we will, even though he isn't constrained by money. When looking for teachers across a broad range of subjects, he will have a tough time just figuring out whom to call. Whereas I'm only one degree of separation from any qualified CS teacher in the world.
If we were to start giving bachelor's degrees we'd have to ask ourselves in what way we were better than Harvard or MIT. Those schools have, respectively, about 350 and 150 years of experience doing what they do. We would be tuition-free but so will they in the long run (I think). The guy below who said that I was an egomanaic is mostly right. I don't like to be involved in something unless it can be the world's best. I know how to make the world's best post-bacc CS program. But I don't know how to build a great college that will do all the things that need to be done for 18-22 year olds.
A guy with an MD or a mother returning to the workforce after a bunch of years home with some kids might not want to subscribe to our brand of fanaticism. And they won't really be immersed in our culture as you suggest. More than half the faculty are not employees of ArsDigita (we have our share of CS PhDs but we've got work to do so we're also hiring CS PhDs whose first love is teaching).
We attack the recruiting problem with little sports cars and other, more immediate, incentives. But ultimately the best thing that we have to offer people is the guarantee that they'll be working with other smart folks (our Pasadena office was started by four Caltech PhDs) and that they'll be free to build great things without meddling from clueless managers or designers.
I think that the reason colleges do 4 courses in parallel instead of intensively is because it leads to maximum admin convenience given (1) full-time faculty, and (2) students in different majors or on different schedules. Note that MIT was built during the Industrial Revolution and works sort of like a facotry. The professors are like machines. They are nailed to their classrooms. The students are like the widgets. They move from machine to machine on conveyors (corridors). Compare to Oxford University, started during the time of guilds and apprenticeships. They have tutors.
1300 on the SATs might be good enough if you took them before the rescaling. Your HS career is certainly better than mine! (see photo.net/samantha/ for an account of my high school achievements). Anyway, we don't stop people from taking the SATs again and crushing them like a bug.
Why do we focus on SATs? MIT has a huge admissions bureaucracy that does a great job but costs more money to run than our whole little school. So we needed some criteria that would yield us a high-quality student body but that wouldn't take up a lot of time to administer.
... and yet here is one of the faculty members who tortured me in grad school, coming into slashdot anonymously. It hurts.
If I were intelligent or level-headed, I would have gone to medical school like my brother Harry.
But rest assured, ArsDigita University students won't have to suffer with me all year. I'm only teaching one course (though I might be a TA for some of the others).
(Note to the rest of you: before you start a little online community like photo.net, be aware that you might get "a few dozen" unsolicited emails over the years from someone like this anonymous coward. It is one of the great joys of being a non-commercial Web publisher.)
Oh yes, the most interesting question you raise is actually answered in a cut-and-pasted excerpt from my Tcl for Web Nerds intro. Why weren't there any good Lisp-based Web servers in 1995 (or now)? Lisp programmers were too busy congratulating themselves for being smarter than C programmers. So the C programmers sat down and figured out what the real problems of publishers and end-users were and came up with practical stuff like AOLserver in 1994, Apache/mod_perl (1998 or 1999 by the time all the db connection pooling kinks were worked out?), etc.
Engineering excellence doesn't mean having a fancier system; it means having a system that solves the users' problem better faster and cheaper.
Am I bitter? No worse than Medea...
(I'm also confused by the person who claims that pgreesnpun isn't me. I am me, dammit!)
Most of the hackers at ArsDigita (including moldy old me) have built Web presentation layers in Perl, Tcl, and Java. We agree with you that Java isn't an improvement over Tcl for merging templates with db data. On the other hand, since the presentation language isn't core to our toolkit, why not make our data models and abstractions available to those who've got big Java libraries (and, okay, the ArsDigita investor in me says "to those who have big budgets")?
Why didn't AD select a Web server that uses Lisp? There weren't any back in 1995 that could talk to an RDBMS. We wanted to build applications to solve real users' problems, not get mired in a tools debate. So we built apps instead of tools. SAP was written in COBOL initially. I'm sure that they didn't like COBOL but they wanted to solve the ERP instead of the language problem so they did (and now SAP is the world's #2 market cap software company, after Microsoft).
Why was I snide? I'm sick of hearing arguments about languages and development tools. It is what gives programmers a bad reputation for being losers.
A few MIT things? The operating system. Time-sharing. RSA encryption. TCP/IP and a lot of earlier network stuff. Computer algebra (and the Lisp-based Macsyma, which is still the most useful computer algebra system (though I admit that Maple is easier to install and use)). As for Lisp, a lot of language bigotry caused Lisp programmers to waste time arguing and reimplementing Lisp instead of building apps. But some good apps were built, esp. in the areas of electronic design and automated mechanical design. A lot of systems used Lisp as an extension language, e.g., the Interleaf publishing system and the AutoCAD drafting system. And a lot of the good ideas from Lisp made it into Java.
But the bottom line is that tools per se won't make much difference. At MIT, we teach our students to generate entire Web sites (including all the scripts) from machine-readable specs. Instead of arguing over tools and maybe getting a 3X productivity improvement (e.g., from using Apache/Perl instead of Netscape Enterprise Server/Java), they are writing programs that write programs. They'll get a 1000X productivity improvement over even the best Perl coder even if their ultimate target language is EDSAC machine code.
The future of Web development is high-level specification and automatic generation of sites. This is what we try to teach students, where we do at ArsDigita when appropriate for a service, and where we are pushing toolkit development.
We'd love to be accredited but it is tough when ours is the only program of its kind. Accreditation doesn't make much sense without a basket of schools teaching the same sort of curriculum.
We don't really want to teach people without a bachelor's degree. If someone doesn't have a bachelor's, he or she can simply apply to MIT and get a regular SB in EECS!
Fear not! As noted below, we're only going to do the first course in calculus from MIT (18.01, which is a few weeks of differential calculus and then the rest of the term on integral calculus). The last thing a CS nerd needs is Div, Grad, or Curl. Nor do they need to know how to use a Frobenius to approximate a solution to a differential equation!
:-)
And if the students don't learn calculus, they can always become lamers like me and use Maple or Macsyma