Recent blog entries for chalst

Abstraction and security
pedro writes: most people don't think about the inherent security cost of using abstractions.

I wrote an LtU story, The irreducible physicality of security properties; key point related to abstraction: "Security is non-modular: Programming languages and software engineering practices can ensure that software possesses properties helpful to security, but the properties are only meaningful in the context of a strategy to ensure a computer system satisfies its security policy".

Recentlog
hypatia: Congratulations! About the thesis-pregnancy conflict — it's maybe a good thing: the nausea, &c shouldn't last for long, everyone expects you to get held up, and a few mothers have told me that the middle months of the preganancy were one of the most pleasantly productive times of their career. So don't worry, and good luck!

hypatia (again): Thanks. I think this may be useful for one of my clients. A good way to think about the big idea behind alternative talk structures is in terms of what story does it tell. This bridges the gap between thinking about what ideas you what the audience to take away and what kind of things you need to say to them and in what order.

Loser SEOs
I see an account topranker spamming topranker.in, an SEO company. This is really sad. I presume that it is the company itself using such a pathetic tactic to gather organic links, though I imagine it could be a rival trying to make their competitor look lame.

I've sent them an email, on the off chance that they might respond.

Postscript — And another one: catvinslate spamming lucentinfotech.com, a web development company that offers SEO services. And also based in India. I guess it's not obvious to every web- service industrialist why this behaviour makes an SEO company look like they don't understand SEO very well.

I'd link to create nofollow links to these, but the Advogato html munger drops rel tags. I think this is a mistake for non-observer accounts, but maybe there are rel tags that we should discourage.

Who benefits from writing quality?
apenwarr muses on the value of professional copy- editing(sic), and writes:

After hanging out on the Internet for a while, you almost stop realizing how low-quality the content is. I'm generalizing here, but I include this very journal, and in fact, this very article, in my generalization. I think I'm a pretty decent writer, but there's no doubt in my mind: a professional editor can help you say the same thing, only better. ... The end result is odd: reduced publishing costs should leave more money for editing and fact checking. Instead, people think those costs should drop at the same speed, which is unreasonable unless you cut quality.

and ends by asking of would-be readers: Would you pay more for quality, edited work? Really?

I think that this question is posed to the wrong class of people. When I buy a book, I will consider my money ill-spent if the writing is poor, and am quite likely not to read much of it. So there, yes, I would certainly pay more for an adequately edited book than for an inadequately edited book. Lots of other readers seem to think so, which is why we are not seeing quite the same economic pressure on book copy-editors than on newspaper copy-editors [1].

But for most classes of writing, it is not principally the reader who loses out from poor-quality editing, but the author. If readers of a weblog find it hard going, they will unsubscribe. The reader will not think twice about it; the author will be asking themself: why doesn't all that hard, content-creating work attract readers? If you want to attract more readers, better writing quality is important, and indeed the copy-editing mailing lists do receive reports from editors who have weblog-editing gigs, likely because edited weblogs attract more organic links, and so more revenue-generating traffic.

And for your book? Lousy editing attracts lousy reviews, which hurts sales, which hurts the author's reputation and the publishing house's sales (relatively few authors are motivated principally by royalties). Just where the most profitable mean avoiding both overly-injured sales and unaffordable editing costs is, the publishing industry has yet to figure out.

[1]: It says a lot when John McIntyre, former president of the American Copy Editor's Society, loses his job at the Baltimore Sun.

Why male managers find it easier getting to the top
From an FT article, Getting women on board by Alison Maitland:

Male managers receive more specialist training for promotion than female managers from the start of their careers and benefit from more support in new roles, an international study by DDI, a talent management consultancy, reveals today.

Starting at junior management level, 28 per cent more men than women receive specialist development via high-potential groups or “talent pools”, according to data from 10,000 business leaders in 376 organisations. At executive level, 50 per cent more men than women get such help. Women, who also receive less support in career transitions such as promotions, may not be aware of discrimination, as selection processes and mentor schemes are “shrouded in secrecy”, says the report, Holding Women Back.

This suggests that the "slippery pole" is a better explanation of low representation of women at executive levels than the "glass ceiling". The same thing goes on in academia; I'd be interested to see independent investigation into this latter phenomenon.

Recentlog
davidw wrote, after noting the poor click-through-rate for his LangPop article's Google ads: Not that I'm really complaining, I did langpop.com for fun, but it's a good reminder that it's a lot better to produce and sell things to a more general audience. Programmers expect, and largely get, most everything for free - me included..

I think it does illustrate that it is not so easy to sell to hackers, but not that trying to is not worthwhile. I read the LP story on RSS, so am the sort of reader who only occasionally sees such ads. After the above post, I looked at davidw's site and was served the following ads:

1. Gratis Website Optimizer
   Testen Sie, welche Zielseite die meisten Conversions erzielt. Hier!
   www.Google.com/WebsiteOptimizer

   Interesting for me, but know the product, thought I knew what the English version of that page contains.

2. C Language Community
   Share knowledge and solve problems with other IT pros.
   C.ITtoolbox.com

   Sounds crap.

3. SPARK Pro
   High Assurance by Design Software development environment
   www.adacore.com/spark_pro

   Interesting for me, again know the product, and thought I knew what that page contains.

I clicked through each, just to see if I was right about the advertised content. And I was.

This is consistent with the alternative hypothesis that hackers tend to find out about things they are interested in before marketing guys put together their advertising campaigns. I'm guessing there may be a niche for a text-ads service catering to hackers marketing their content to other hackers.

The diabolical confederation of scholars
ncm: Today I was going to be so busy that I couldn't compose an Advogato entry, but then you wrote:

Continuing to follow up, note this from the conclusion of the BitC paper: "It is noteworthy that none of this effort was deemed fundable by the National Science Foundation (which is to say: by the academic programming languages community)." Mm-hmm. All BitC needs now is destructors and exceptions. If they can avoid any analog of "finally", then, they're home free. :-)

I asked Jonathan Shapiro about what he made of the complaint of Bjarne Stroustrup's that you had reported; I should think you would be interested in his response, where he says [t]he effect that Bjarne describes is real, but I don't think it's a matter of bias on the part of academics.

Now I guess I have to spend more time dealing with outstanding points! On Friday, you wrote:

I have the advantage of having no desire whatsoever to preserve existing Haskell code, nor Haskell coding conventions, nor favored Haskell idioms, nor even laziness-by-default. I'm talking about a language that preserves the fundamental strengths of Haskell-like languages (which does not include their built-in Lispy data structures), adds the fundamental strengths of C++ (most particularly the destructor), and stirs in enough practicality to make the language industrially useful.

I'm guessing that BitC is a better starting point for the kind of things you think might be good about Haskell than Haskell is, especially if you have doubts about the value of monastic discipline. So I'm pleased to see you say earlier today: Following up, Bitc looks like it could become quite interesting.

Having plugged BitC, I still think that Haskell is the world's greatest language laboratory for many reasons. Monastic regions, in particular, could be of interest because region inference is, in essence, rather like using type inference as a basis for automatic inference of constructors and destructors. Add some decent instrumentation to this, improve Haskell's built-in runtime costs, and perhaps you might find the result useful.

Haskell region management
ncm wrote: Haskell lacks the facilities to enable programmers to code and encapsulate resource management.

There's been some work done on the region management of state under the banner of monadic regions, and Google tells me that there's an implementation of the idea for ghc in the Hackage code repository, Ivan Tomic's IOR. I haven't looked at the details of that package, but in principle it allows you to manage certain explicitly declared state, which might be exactly what some programs need. It doesn't help you with managing the undeclared resources, part of Haskell's hairy runtime cost, so it doesn't make Haskell suitable for the kind of system programming that BitC is designed for.

Add the ability to parse C header files directly, and call functions declared there, and most of your standard library collapses to trivial wrappers.
C++ pays and pays for it, but the great advantage C++ gains from simply including almost all of C is that there's nothing foreign about interfacing to C. In the paper introducing c2hs, Manuel Chakravarty says:

SWIG works well for untyped scripting languages, such as Tcl, Python, Perl, and Scheme, or C-like languages, such as Java, but the problem with typed functional languages is that the information in the C header file is usually not sufficient for determining the interface on the functional-language side. As a result, additional information has to be included into the C header file, which leads to maintenance overhead when new versions of an interfaced C library appear.

The two FFIs for Haskell, c2hs and Greencard, take opposite routes. Greencard, like SWIG, includes extra info in the header file, Greencard needing quite a bit more than SWIG, while c2hs provides an interface to allow Haskell wrapper code to be written to bridge the information hole, but of course that introduces another kind of maintenance overhead.

What you want doesn't seem to gel with Haskell.

Runtime cost 3
graydon: The runtime cost of a particular VM doesn't end with the bag of goodies that it may or may not provide; VMs also impose assumptions about what the target architecture looks like, and provide constraints on the freedom of compiler writers on how they realise their language. I've generally looked at these infrastructure costs as being the more important question to ask about a VM. They seem to be the reason why FP implementers have mostly been lukewarm about the LLVM.

It's a fair point that there's a lot of hard work involved in providing tasty bags of goodies in the runtime; the effort that has been put into duplicating the JVM's runtime outside the JVM (i.e. Gnu Classpath, though I guess the free JVM's can be mined for suitable code) lags behind what Hotspot provides. I tend to think that, while this matter to how attractive a programming environment a language provides, little of it is relevant to the core language. Gnu Classpath provides a nearly adequate basis for implementing Clojure on LLVM for my purposes, although perhaps not for johnw's

I had just run into VMKit myself as well, and I'm curious. It says it doesn't support "mixed interpretation/compilation": is that an architectural limitation, or unfinished work? It certainly looks interesting.

Postscript: ncm wrote:

I find it a little bit ironic, under the circumstances, for Haskell to be a pig, considering that the Haskell/Caml family of languages has, potentially, much less use for garbage collection than other functional languages, and could, with (what are really quite) minor language changes, live with runtime support no larger than C++'s.

I'd like to believe this, but I don't see how this can be true. The dominant programming paradigm for Haskell, in particular, encourages heavy use of higher-order functions, which, at least according to the obvious interpretation, requires very heavy heap usage for allocation of closures. Techniques such as substituting region inference for GC might improve predictability of resource usage, but they seem to do so at the expense of efficiency; perhaps FP allows one to choose which runtime costs one has, provided they are heavy.

Runtime cost 2
graydon: Good to see you around on Advogato! I think what I wrote was unclear: johnw has been very enthusiastic about Haskell recently, and in the light of that it seemed strange to bash Clojure because of the kind of run time commitment it has.

I certainly see the whole question of the grasp of and control over the runtime as being the big issue between high- and low-level languages, and I've been pleased with the direction that research into high-level languages has taken in this respect.

I disagree with only part of what you say: I don't see the sharp distinction you seem to make between LLVM and JVM. Certainly the LLVM is a minimally abstracted VM in a way that the JVM is not, and which does not set out to solve many of the problems that the JVM does, but there is an obvious respect in which they accomplish the same kind of task. Equally it's not absurd to talk of compiling Clojure against non-JVM architecture, just as gcj compiles Java to the gcc runtime.