SERVING THE QUANTITATIVE FINANCE COMMUNITY

  • 1
  • 2
  • 3
  • 4
  • 5
  • 49
 
User avatar
lballabio
Posts: 983
Joined: January 19th, 2004, 12:34 pm

Three years from now: can you predict the Software Landscape in anno 2018?

August 31st, 2005, 12:32 pm

QuoteOriginally posted by: exneratunriskdocuments which look like text books but work like programsLike this book? It uses facilities such as those available in Haskell (see here) in order to be both a book and an executable program. It was published in 1992...And as mentioned by Athletico on another thread, Literate Programming can deliver this as well.
 
User avatar
DominicConnor
Posts: 11684
Joined: July 14th, 2002, 3:00 am

Three years from now: can you predict the Software Landscape in anno 2018?

August 31st, 2005, 12:53 pm

A paradigm I'm confident that will be important is some form of SIMD processing.>99% of quant programming is currently of the form:Get a numberScrew with it until it gives inGet other number, somehow related to first numberScrew with that until it is moreproperly related to first numberGet next numberSIMD, Single Instruction, Mulitiple Data, means doing the "same" thing to a set of values.An obvious first step is matrix operations, but there's a lot more to it than that.Even with today's rather awkward technology you can get a factor of 10-20 by radically recoding your numerical methods this way. (that's a factor of 10, not 10%).Is quite hard, not least because currently we don't even have debuggers for this stuff.However it works on most current hardware, and better stuff is coming.
 
User avatar
exneratunrisk
Posts: 3559
Joined: April 20th, 2004, 12:25 pm

Three years from now: can you predict the Software Landscape in anno 2018?

August 31st, 2005, 1:47 pm

o.k. I withdraw the big word paradigm shift.But if we want to grow "plants", like literate (declarative, descriptive, task -oriented,...) programms or genetic/evolutionary programms, SIMD based programs,..... To predict the SW Landscape: What will be the best "substrate" to develop, deploy, grow and harvest them? Not the web?
 
User avatar
DominicConnor
Posts: 11684
Joined: July 14th, 2002, 3:00 am

Three years from now: can you predict the Software Landscape in anno 2018?

August 31st, 2005, 2:28 pm

Another paradigm shift I see is the idea of financial instruments as programmes. Already this is true to a certain extent.Averaging swaps, path dependant options, and the formulae embedded within the definitions of a variety of exchance traded and OTC instruments will continue to grow in complexity.Already it is a challenge to price a credit instrument when the "risk" is effectively determining the behaviour of the program that keeps it hedged and marked to market.The contracts that constitute the "reality" of an instrument is already a primitive programme, even when they have no maths.They declare their variables (the party of the first part, etc)They have procedural code, as well as exception handling.I have not seen contracts with Excel sheets within them, but they're coming, and I've been shown things where the model and it's drivers was integral part of the thing being sold.This will cause a step change in models. From jump/diffusion we will have to grapple with the stochastic properties of finite state automata. I've played with this, and and it's so far outside my understanding that I can't even tell if the rubbish I get is really crap, or something very deep. (Betting is on crap).
 
User avatar
N
Posts: 2808
Joined: May 9th, 2003, 8:26 pm

Three years from now: can you predict the Software Landscape in anno 2018?

August 31st, 2005, 2:42 pm

I don't think we'll see any change is the software landscape over the next 5 to 10 years.Maybe we'll see more multi-threading because of hardware changes, but that's it.
 
User avatar
Cuchulainn
Topic Author
Posts: 59712
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Three years from now: can you predict the Software Landscape in anno 2018?

August 31st, 2005, 2:57 pm

QuoteOriginally posted by: NI don't think we'll see any change is the software landscape over the next 5 to 10 years.Maybe we'll see more multi-threading because of hardware changes, but that's it.Great, old Cuch can stop learning now? P.S. Do you know anything about gravity wells?
Last edited by Cuchulainn on August 30th, 2005, 10:00 pm, edited 1 time in total.
 
User avatar
lballabio
Posts: 983
Joined: January 19th, 2004, 12:34 pm

Three years from now: can you predict the Software Landscape in anno 2018?

August 31st, 2005, 3:36 pm

They're usually created by someone^H^H^Hthing being dense.
Last edited by lballabio on August 30th, 2005, 10:00 pm, edited 1 time in total.
 
User avatar
Cuchulainn
Topic Author
Posts: 59712
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Three years from now: can you predict the Software Landscape in anno 2018?

September 1st, 2005, 7:26 am

QuoteOriginally posted by: lballabioThey're usually created by someone^H^H^Hthing being dense.?Non ho capito
 
User avatar
lballabio
Posts: 983
Joined: January 19th, 2004, 12:34 pm

Three years from now: can you predict the Software Landscape in anno 2018?

September 1st, 2005, 9:01 am

I see---obviously a bad joke. Gravity, mass, density...
 
User avatar
Cuchulainn
Topic Author
Posts: 59712
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Three years from now: can you predict the Software Landscape in anno 2018?

September 1st, 2005, 9:20 am

QuoteOriginally posted by: lballabioI see---obviously a bad joke. Gravity, mass, density...http://www.wilmott.com/messageview.cfm? ... adid=30849
Last edited by Cuchulainn on August 31st, 2005, 10:00 pm, edited 1 time in total.
 
User avatar
alohashirt
Posts: 60
Joined: July 14th, 2002, 3:00 am

Three years from now: can you predict the Software Landscape in anno 2018?

September 2nd, 2005, 2:50 am

Dr Ben, Cuchulain, I think you're mistaken. The OMG have a track record of producing overcomplex, bureacratic, ugly standards that promise the world and ultimately disappoint. If you think about Corba, UML, XMI, MOF they all share two things:1) They are fat and ugly.2) They are products of the OMG.If you look at the standards that have appeared over the past ten or twenty years you will see two distinct camps: 1) Standards that were "harvested" from existing innovative technology, typically developed by a couple of people. 2) Standards that were created as a foundation by a larger group (committee) that identified a hole, created a standard with a reference implementation following shortly afterwardsIn the Java world that I know, examples of the first type are the JDBC standard, the Servlet api, and the JDom standards which standardized the Weblogic Tengah database driver and the Sun Java Web Server, and JDom. Examples of the second are the EJB, the JMX standard and the JAXB standard.Any of the first three specs can be read and comprehended in one sitting. Their implementations solve a very specific problem quite neatly. The second three are butt ugly, difficult to comprehend and imperfectly solve overgeneralized problems. I'd argue that the EJB standard in particular has been responsible for more project fail;ures than any other in the history of software development. Whilst I can't see executable UML matching EJBs record I am extremely skeptical about its utility. Ultimately the end point of development is code. Visual pportayals of code are extremely useful and are certainly underutilized. The best bang for buck here is often imperfect scrawls on a napkin whiteboard or back of an envelope that show ambiguous bozxes and lines and have a shelf life of about three minutes in a technical discussion. The step up from this of using UML to diagram one p[art of an existing system, perhaps to justify to management whny something took N months can be useful. But of course the challenge here that cannot be automated is what to leave on the diagram and what to omit. It's a little like when Rodin was asked "How do you sculpt?" and replied "Simple. I choose a block of marble and chop off whatever I don't need." Thjere are different personality types who choose programming as a profession. One type who are often pretty good are what I think of as the "black and whiters" who see programming as a refuge from the muddy inconsistent world of people and like the reassuring certainty of syntax. For them MDA plays into their desire for unambigous requirements but, as this view of technology is incomplete, it will never be the silver bullet they are hoping for. Ultimately programming is an antisocial social activity. When we write code we're beginning a conversation with the next N programmers and we forget this at our peril. Imagine inheriting a system that is written in executable UML. Yes in theory UML can completely depict a system - and in theory men could carry children in their abdomens but I'd happily bet $10k that men won't be queueing up to do this in my lifetime. MDA would eb useful if the world were describale by UML. It isn't.Stroustrup started work on "C with Classes" in 1979 and Gosling began work on "Oak" 12 years later in 1991. I think Java has really just become useful now. I think the next wave will be higher level languages that use dynamic typing and are less verbose than Java and less arcane than C++. I think the obvious candidates for language of the next decade are Python, Groovy and Ruby - all ric h typed OO languages with some functional characteristgics and good sized class libraries. The minor technicaL advantages that C# has over Java seem outweighed by the lack of third party libraries or choice of developer tools. I expect that C# GUIs will be fashionable for some enterprises but eventually rich clients will eb written in a more concise scripting language. I also think that developer tools will become more important as we need to leverage larger and larger code bases. I do not expect to see fruit baskets returning to the workplace in the next ten years.Enough ranting.
 
User avatar
DominicConnor
Posts: 11684
Joined: July 14th, 2002, 3:00 am

Three years from now: can you predict the Software Landscape in anno 2018?

September 2nd, 2005, 7:19 am

Excellent points by Alohashirt.Corba is about the ugliest programming framework in the history of the world.I do slightly part company on dynamically typed languages as being mainstream any time soon. Not because they are bad, but because they stress out the average programmer too much.But if they evolved a bit, I'd revist them.Never liked Rexx.That was very much one man's vision.
 
User avatar
Cuchulainn
Topic Author
Posts: 59712
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Three years from now: can you predict the Software Landscape in anno 2018?

September 2nd, 2005, 8:00 am

> Dr Ben, Cuchulain, I think you're mistakenAlohashirt,Good well-balanced post, I don't find it a rant btwIn what area of discourse do do not agree? I reckon the UML stuff??To be honest (and now a bit arrogant I suppose) I use UML as a mechanical engineer uses CAD (design) drawing to design a problem and then map this drawing to a CAM (manufacturing). This is unfortunately is not what happens in generall in IT at the moment (I know, really).OMG: that's a good one and I agree with you. Sometime I will tell you I tried to influence it to adopt a more practical standpoint. No success. The less said the better.UML is good if u use it well. I use it as a design BLUEPRINT instead of goofing around at a keyboard trial and error. It all takes time, that's progress. In 1800 civil engineering was still ad-hoc. By 1880 there was enough knowhow to build San Francisco bridge.
Last edited by Cuchulainn on September 1st, 2005, 10:00 pm, edited 1 time in total.
 
User avatar
Cuchulainn
Topic Author
Posts: 59712
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Three years from now: can you predict the Software Landscape in anno 2018?

September 2nd, 2005, 8:07 am

On a more general follow on question:Is there a trend to get the design right first-time as in chip design, for example (kind of Clean Room idea).
 
User avatar
lballabio
Posts: 983
Joined: January 19th, 2004, 12:34 pm

Three years from now: can you predict the Software Landscape in anno 2018?

September 2nd, 2005, 8:47 am

QuoteOriginally posted by: CuchulainnIs there a trend to get the design right first-time as in chip design.I see two difficulties with it:a) I might be wrong, but my impression is that any non-trivial software project is more complex than chip design. Also, it has complexity at several levels, e.g. both at the bird's eye view level and in the details (where the devil often is.)b) getting the design right first-time requires having precise specifications up-front. This hardly ever happens, as specs tend to evolve in time. Designing in order to allow for change is a most difficult task---especially as changes often takes unexpected directions...Luigi
ABOUT WILMOTT

PW by JB

Wilmott.com has been "Serving the Quantitative Finance Community" since 2001. Continued...


Twitter LinkedIn Instagram

JOBS BOARD

JOBS BOARD

Looking for a quant job, risk, algo trading,...? Browse jobs here...


GZIP: On