Jul 172005
 

I’ve done a very minimal amount of reading up on the Internet2 project, but I’ve been thinking on it quite a bit. It seems that the underlying motivation is to clean up the usage model for on the wire communication. That is to say that the way two computers communicate in the current world is open and uninhibited except by internal constraints within a LAN, WAN or WLAN, most commonly firewalls. Packet transmissions that get passed firewalls are free to roam the Internet however they please.

Another interesting thing to include is the obvious UDP restrictions imposed in nearly all corners of the Internet, especially between two LANs. Internal LANs use UDP freely because it provides the a simple broadcasting mechanism. Outside those LANs, UDP packets are often dropped immediately. This is almost expected considering the breadth of the Internet and the less than ideal semantics of UDP.

What’s the end goal? I think the main idea is a way for computers to communicate that restricts SPAM, hacking and phishing while allowing people to send email, use applications and get their work done, otherwise what would be the point, in works right now except for the 300 SPAMs I get each day.

I have an email certificate. I think everyone should have one. Hell, I think every computer should have a certificate from an authority. Just roll it into the price of the machine and hard-wire it to the motherboard. Yeah, I’ve heard the old, “the government can track you,” and “Microsoft can watch you,” arguments. They just don’t seem to bother me. If everything had a certificate, the government and Microsoft and any other company would have to have computers installed everywhere with the ability to man-in-the-middle ever packet ever sent out to any place. They miss a single packet, then for the rest of eternity both computers know about each other and have each others certs and once you encrypt the packets with those, you are set.

Jul 162005
 

I read an interesting article by Sergey Dmitriev of JetBrains on LOP and DSLs. Here’s the link

http://www.onboard.jetbrains.com/is1/articles/04/10/lop/

I had some initial thoughts:

1. What is the learning curve for these DSLs and each LOP going to be?
2. How many LOP instances could we expect and how many of us will need to learn them?

The main reason these come to mind is the constant battle to standardize in the Java world or at least “de-facto standard”ize. This stuff is important because I can put Spring and JSF on job descriptions, interview questions and training program and be able to get engineers into the code much quicker. Are DSLs so simple that the learning curve is reduced enough to make them benefitial? Or are they increasing the number of days (weeks) (months) I have to wait until a new hire can work in the code making them less beneficial in large applications with hundreds of developers. Likewise, what is the breadth of DSLs required for large enterprise applications and what is the trade off in creating new ones for new features/applications as opposed to the use generic languages when considering time to market and business development ideals?

Something I’ll have to think about some more.