Keith Bierman's Weblog

The importance of a solid paper trail... From Kirk/BSD
(2004-05-24 13:04:00.0) Permalink

Does this mean that "Discovery" results in every laptop being turned over to the Court? At Microsoft?
(2004-05-24 12:32:11.0) Permalink

Some old, but good books When Gravity Fails as the link provides a pretty good review, I'll leave it at that. George's writing never got all the "airplay" I thought it ought, but such is life.

For most of my life, I've walked around with a paperback in a pocket, to while away time in lines and such. One of the nice things about modern PDA's is that they'll fit at least one book. A library of cheap or free pulp stuff can be found here. I had never gotten around to reading the Doc Savage novels; certainly not great literature, but a classic of the genre.

My favorite example of the genre is the E.E. Doc Smith novels. A review. I could have sworn there was a lovely hardback reissue a couple of years back, but my copy never came though (I was expecting it as a gift) and I've lost track of who did the reissue.

A good source of free (as in beer) online novels is the Baen Free Library. I enjoyed rediscovering some old authors and discovered  a few I hadn't come across (and, naturally, bought dead tree versions of some of their stuff). Proof that some publisher's do "get it".

A couple of books with unlikely heros which I really enjoyed: Vlad Taltos (ok it's a series) and A Night in Lonesome October. In the former, the protagonist is a professional assiasn and nice guy, and in the later it's Jack the Ripper's Dog. Jack being one of the few to stand between the world and obiteration (who would have thought that he was one of the Good Guys?)

In terms of serious books, a fairly famous one that's a great read is Guns, Germs and Steel and also fairly well known is Punished by Rewards

Oh well, enough of a random walk through my dusty library.



 

(2004-05-23 22:02:44.0) Permalink Comments [1]

More from A. Tannenbaum A followup to "..considered harmful" (my ripoff, not his).
(2004-05-22 23:50:44.0) Permalink

Tocqueville considered Harmful This writeup is a classic. A wonderful read; a first hand review of important history, and topical.

Sometimes they make valid points ... but you've got to wonder at the gross incompetence ....

(2004-05-20 21:35:23.0) Permalink

The importance of being honest about one's corporate afflilations. See this writeup about what some JBOSS employees are purported to have been up to.
If they'd posted honestly, they might have been equally effective (it's the quality of the argument, not where you post from) without the possibility of serious fallout afterwards.

And, of course, posting under proper attribution makes one a bit more careful about how one makes one's arguments ;>

(2004-05-18 15:27:32.0) Permalink

gas boycott A well intentioned but naive friend of mine sent me the latest incarnation of this urban legend
Also a good writeup can be found here
A related bit of nonsense.

For a good treatment of the underlying issues see as well.
(2004-05-14 09:30:21.0) Permalink Comments [1]

Reinvention of newsgroups.... Well, not quite. But I stumbled across this posting which harks back to my first blog posting (internal, later reflected as posting #2 here).
(2004-05-12 23:04:33.0) Permalink

BitKeeper makes Linus 10x more productive! I've known Larry McVoy for a long time (faint recollections of a practical joke that went slightly awry...). While I haven't used BitKeeper for anything significant myself, I've followed it's progress from afar. This article about how it's made Linus 10x (OK, Larry only claims 2x, but the quoted anecdotal result is 10x) more productive is intriguing. How many software tools actually deliver that much productivity enhancement?

Part 2 is now available.
(2004-05-11 17:58:42.0) Permalink

A couple of pointers The first is a little essay on what make s good blog.

The other is a video lecture on computer architecture by Bob Colwell. It is in windows media format, Xine seems to play it fine on JDS. As one of the principal architects of Intel's P4 and a Multiflow survivor he certainly has good war stories. I don't always agree with his analysls, but he is a terrific speaker and one ought not ignore him (had Intel paid more attention, IA64 might not have turned out as it has).
(2004-05-11 08:17:43.0) Permalink

Object-Oriented Programming In the beginning, there was SIMULA. One of it's original creators was Kristen Nygaard 
I never had the pleasure of meeting Kristen; but from the many lovely memorials that clearly was my loss.
(2004-05-08 23:51:36.0) Permalink

Canceling CPU's at the last minute There seems to have been an amazing amount of confusion kicked up by Sun's cancelation of a couple of chip projects late in their development. Now it may be Intel's turn.


In case anyone is confused about Sun's situation, this may help. You'll have to look elsewhere for details about what Intel is actually doing ;>

An Open Letter from David Yen, Executive Vice President

Scalable Systems Group, Sun Microsystems

Over the past week or so, some articles have been published carrying the news that Sun has discontinued its SPARC® development projects code named “Gemini” and “Millennium”. A number of analysts and industry insiders have praised this decision, correctly interpreting it as a commitment by Sun to bring our next-generation Chip Multithreading (CMT) technology processors and systems to market as quickly as possible. However, some writers and a few competitors have suggested that these actions signal Sun's decision to reduce its commitment to SPARC.

I want to use this first opportunity to address you as the Executive Vice President of the newly-created Scalable Systems Group, that we are unwavering in our commitment to the SPARC/Solaris^TM platform. SPARC and Solaris continue to be our foundation for delivering premier solutions for secure, reliable, and powerful enterprise computing. We made the decision to discontinue the Gemini and Millennium projects in order to better focus our SPARC development efforts, and to hasten the date at which we can deliver far more powerful SPARC/Solaris platforms based on new Chip Multithreading technology.

What do our recent decisions mean for our roadmap? What do they mean for our customers?

We have a very strong stream of UltraSPARC® IV and UltraSPARC IIIi upgrade projects scheduled for delivery over the next several years. These new processors will keep our SPARC-based systems highly competitive, and will provide our customers with a continuous stream of performance enhancement opportunities. Our latest generation of Sun Fire^TM hardware, introduced in February and based on the UltraSPARC IV processor, has already set new marks for OLTP and database applications. As a result of successes like these, SPARC today is by far the highest volume 64-bit processor shipping for enterprise servers worldwide.

Moving forward, we will focus our substantial development resources on Niagara and Rock processor families, which incorporate entirely new pipelines, optimized for their targeted workloads. Both of these SPARC designs are based on powerful CMT principles, and fully support our revolutionary SPARC/Solaris Throughput Computing initiative. The Niagara and Rock processors are already well along in development, and our intensified focus will enable us to get them to market as quickly as possible. These network and data facing processors are expected to deliver an order of magnitude gain in throughput performance. We will deliver these powerful new generations of CMT-based SPARC processors while continuing to protect our customers' investment in software by maintaining binary compatibility.

In closing, I am confident that the future of the SPARC/Solaris is better and brighter than it has ever been in its already extraordinary 17 year history.

David Yen

Executive Vice President

Scalable Systems Group

(2004-05-06 23:28:35.0) Permalink

Performance still matters! James Hague  makes the case that machines are so fast now that optimization doesn't matter. It's an argument frequently made (by everyone except computer vendors, whose potential customers tend to make it clear that performance matters with respect to their purchase decision ;>).

For people working on many problems, it's true that optimization is something best left to automation. It's why so much work takes place in scripting languages, and the like. And it is a GoodThing.

But folks doing serious near real time data analysis (CAT scanners, MRI, etc.) certainly have a very different set of needs. So too the folks doing weather forecasting (it does fairly little good to forecast tomorrow's weather, tomorrow). For many problems computers are still orders of magnitude too slow to do a really good job. Careful performance analysis allows for better solutions, in a cost effective fashion.

If you really don't believe performance matters, consider that the bridge you drive over, the car which you expect to protect you in the face of a collision, and the airplane that takes you to vacation were all designed and simulated by software whose performance darned well ought to have mattered.

Because lower performance translates directly into less safe.
(2004-05-05 22:10:03.0) Permalink

A cute macromedia flash test case Having just upgraded my laptop to Java Desktop System v2, I wanted to make sure that flash worked correctly (sometimes it was necessary to reinstall the flash component on JDSv1).

Here's a site which provided me with some entertainment value as well as a test case.

JDS v2 is working quite well.
(2004-05-05 15:43:47.0) Permalink Comments [1]

Some ruminations on usenet, vs. blogs vs. email Inspired by Danese  and AlanC and of course Alec I was motovated to do some googling of myself, and pondered electronic communities a bit.

Interestingly Google inconsistently finds my earliest posts. Indeed, one  from comp.benchmarks which made it into a book is only showing up about 1 in 10 google searches (or a9.com searches) Here's the book (I believe that Amazon requires an account to search within the book, search for "bierman") Measuring Computer Performance

In any event, my earliest postings date from 1988.

At first Sun had no identifiable policy against it (or at least none that anyone pointed out to me, including my manager). Then there was an era where it was restricted, but I had a "note from my Management" and after a year or so of having to waive it, the netpolice left me alone (and now there's an open policy ;>).

But onto the ruminations themselves:

• In my earliest computing days, "community" meant hanging around the card punch machines or printers. There were newsletters and the like.

• Then, during the CP/M (and less well known compatible multi-processor TurboDOS) era, bbs's provided various balkanized communities. There was no reliable way to look up old postings (every sysop had different retention policies) and while a given bbs might have various interest areas, information from one bbs was seldom propagated to all others (although fidonet certainly permitted it).
  
• Then I wandered into usenet in late 1988. Corporate internal newsgroups allowed for good community building within the organization, and the greater usenet provided many useful (or entertaining) fora to discover people of similar interests.
  
• Of course, at the same time, an increasing number of email lists provided community support (accessible even to the least computer literate).
  
• Now we have blogs .... I don't know if anyone has done a formal study of the differences in quality of community and satisfaction in the community. I imagine it would be interesting.
  

My personal experience is:

BBS  Aside from the balkanization, using a BBS was very workflow disruptive. One had to dial out, login, read/write and wander to the next BBS. If one was trying to do real work, one needed two modems ... and possibly two processors. History was iffy.

Usenet   Overall a very wonderful setup. Smoothly integrated in the daily workflow (assuming one had decent networking bandwidth). Messages resided on the server, and headers could be downloaded to the reader separate from the the bodies ... so it could be very time and bandwidth efficient. Thanks to various archives (now almost entirely taken over by Google) everything posted remains in cyberspace (some may feel that's a mixed blessing:>)> The primary drawbacks which eroded my (and I suspect many other people's) usage:

1. Chronic flamewars.
   
2. Open to idiots (increasing noise to signal ratio)
   
3. Lack of consistent corporate support (odd, in that the load for a newsgroup was a fraction of that for a large email alias, but that's corporate culture for you).
   
4. Every server has (typically) it's own article numbering, so switching client systems results in either losing messages or having to reread many old messages (and an increasingly itinerant workflow made this extra tedious)
   

Email  As email lists are typically centrally controlled, it is possible to exclude idiots (depending on local list policy). Lists may be moderated. Email is the bedrock for most people's daily workflow. On the downside:

1. Without sever side filtering, or good client sorting, being interested in an increasing number of groups can result in a more interupt driven day (most people can't ignore an ever insistent email client ;>).
   
2. A single large message is replicated in every consumers mailstore (unless they happen to all be on the same mail server, and it's imap based, and well configured...).
   
3. Most are not moderated, and even otherwise intelligent people seem to forget how they subscribed or were instructed in how to unsubscribe. So "storms" triggered by some noise threshold having been crossed, escalate almost endlessly as folks ask the list (rather than the list owner) to unsubscribe, and all the discussion and others confused people trying to unsubscribe the same way...
   
4. While idiots can be banned, one person's idiot is another's savant...
   
5. Of course, no one can count on all copies of a message they've sent being deleted; but neither can they count on being able to read it years later (unless they personally saved it).
   
6. Unlike news articles, there's no expiration time, so postings with a limited "shelf life" (e.g. there are cookies in the breakroom) clutter up reader's mailboxes long after events have made it moot.
   

In bbs's, usenet and egroups, most postings are in response to other postings. So the mode of discourse is endless discussion. As discussions are nominally focused by topic, one may see very different sides of the same person in different groups.Postings may well come from people one whose opinions one doesn't think much of as often, or more often than those one is interested in. The advent of imap makes it practical to change clients @ high frequency without undue effort.

Blogs  While there certainly are some which invite discussion, most seem to end up as series of monologues with pointers to other blogs and items of interest interspersed. Of course, no one pushes a blog onto the reader ... the reader chooses whose blogs to read (or subscribe to via RSS, etc.). Indeed, instead of being organized primarily by topic, the organization is primarily "does $Fred generally have anything interesting to say" . It's less clear to me just what the longevity of blog data is, will postings be reliably available years later like usenet?  It has somewhat the feel of the old BBS systems (you know where to look for the good stuff, and your friends know where you hang out, etc.)

So which provides the best community or community building? My guess is a combination. blogs to provide the input to dynamically created newsletters (via RSS) and probably email (I suppose twiki's might work well too) to provide forum for discussion.

My "deep thought" if there is one:

(2004-05-04 15:45:28.0) Permalink