Discussions

I've used it for years!  Probably since it was something like 0.9 alpha.  I'm glad to see it released to the public.  It is very useful and I like the syntax better than R (which I've been avoiding).  R has far more packages, but Sho is super easy to work with for random tasks and easily integrates with .NET. 

I wanted something like Matlab that didn't cost a fortune and found that MSR had this as an incubation project which met my basic needs

W3bbo said:

spivonious said:

*snip*

There aren't 64 levels, it actually starts at 50 and goes up to 70, according to these leaked documents.

MS pays at the 65th percentile for the role and experience.  This is widely documented on the internets.  As for levels, you NEVER see a developer at less than 59.  There aren't 20 levels.

100K isn't "rich", but you don't have to worry about money.  The benefits are great too (again, not the best on the planet, but it is hard to beat the healthcare anywhere in the US).  Also, that talk is about base salary.  Bonus and stock for good performance can make that much more. 

MS pays for performance.  A Phd who proves they can deliver more than a person with a BS will quickly make much more money then the person with the BS, even if they start with similar salary.  If the person with the BS can deliver more than a person with a PhD, that will be reflected in the pay as well.  As a person who has PhD's and Masters and BS educated employees working for me over the years, there isn't a direct correlation to getting things done and adding customer value. After getting hired into a product group (MSR may be different), your education isn't nearly as imporant as delivering high quality results.  If your education helps you do that, more power to you, but I gaurantee that your education plays no direct role during review time.

jamie said:

MasterPie said:

*snip*

i like that!

 

*edit: ill tell you one thing - if EVAR the weight of ms creative might/ should fall on my shoulders - i would lift it like atlas.

 

pause for laughter....

 

 

edit: anyway...save something - do something .... its not too late... ITS STILL GOING ON

 

 

pps: delete this thread - get on cnn - pledge 200 mil and urge other ceos to do the same   PEACE!

 

*NOTE to ballmer... LEGACY!!!!!!

MS help?  BING! to the rescue.  http://blog.seattlepi.com/microsoft/archives/210083.asp

I hope we may AVOID POTENTIAL MISUNDERSTANDINGS OR DISPUTES WHEN MICROSOFT'S PRODUCTS OR MARKETING STRATEGIES MIGHT SEEM SIMILAR TO IDEAS SUBMITTED TO MICROSOFT.

AndyC said:

kriskdf said:

*snip*

Big O is great in 'theoretical' terms, but in practical terms 'N' is always bounded and as such Big O isn't necessarily a good indicator for real world comparisons of two algorithms/data structures since the effect of constants dropped from a big O calculation can be very significant. Practical things like page faults and the non-linear speed of memory addressing (due to layers of caching) can be, and generally are, far more significant in determining how something scales within realistic bounds.

well...i guess if you make N a constant by giving it a practical bound you can decide which contant has the biggest impact on overall performance.   I work on Bing and N can be very big and continues to grow.  Because N is so big, the BigO is important, as is the constant.  Spending 1 second on a web doc has huge impact compared to 500ms.  Going from O(N) to O(NlogN) is huge if N is bounded at 100B or so.  How much would you need to reduce the constant time to make it worth sticking with O(NlogN) over O(N)?  How about at a trillion docs?

So i guess the conclusion is that BigO and the constant are important depending on the situation.  

PaoloM said:

Nobody (statistically) was using it, so they discontinued the project.

 

Sadly, I'm one - like you - of the nobodies

 

I guess you can use the OPML on other aggregator sites, but I didn't find a really good one so far... maybe it could be an interesting Ch9 project....

i manually added stuff to my.msn.com and it wasn't as good for me.  I sent feedback and then downloaded RSS Bandit.  I'd prefer to keep my.live.com account going.  I like that I can see all the headlines on one screen without scrolling.  Not possible on my.msn.com.  Booo!!!  

magicalclick said:

Sven Groot said:

*snip*

I am not talking about what BigO is for. I have being saying we don't want explosion of compution from the begining.

In real life, Big O is most important for the design and architecture review.  The C is important during code reviews, development, and testing.  If you can't get the right algorithm to scale for your problem, you are doomed.  Once you have the right algorithm, you should use a profiler and measure the best way to reduce the cost of "C".  For school, it seems like the right thing to focus on is BigO since the amount of time spent on performance tuning is really about finding a good balance of meeting real world requirements and managing cost of development.  In school, they don't tend to focus on the best compiler settings for performance, measure unit test coverage, do threat models, run PreFast, or other such tasks, so in most cases, they also don't focus on other optimizations that are constant time. 

BTW, the constant time is sometimes considered upfront in real world applications if it is obviously going to be a problem.  For example, for each N, do you need to go to memory, disk, or network?  Knowing some of these things can change an algorithms design (maybe do work in batches of 1K N instead of individual N, especially if N is going to be very big).

code complete is awesome.  If you are serious about a career in programming, I would also add:

Writing Secure Code.

And a good book (actually...I found training to be useful) on design patterns.  I'm not sure what's the best here.

If you are concerned about the amount of memory the application is using, why don't you measure that instead of the list?  You can compare the process memory to the available memory to avoid paging.  This is really just an approximation (another thread could be eating memory) but it may be good enough?  The advantage is that your memory limit automatically increases on new hardware or decreases if other processes are eating up memory (let's say SQL is running on the same machine).

Depending on your app, i would consider sorting in memory and flushing partitions to disk (then build another partition in memory and repeat).  Then you can just merge the partitions once your are ready instead of worrying about a disk based sort. 

eagle said:

PaoloM said:

*snip*

It never is, but if you want to believe that an experenced Journalist suddenly had the inclination to publish a story on Microsoft re-branding search independently of any impute from a Public Relations professional you go right ahead.

If your anywhere near a computer why don't you try an MSDN search on http://bing.com and let us know how you feel it compares to Google. thank-you

I use bing to search msdn quite frequently and know that bing has solid coverage of that domain (bing dev here....lost my Microsoft post background thingy in the lastest C9 upgrade apparently).  Are there specific queries that underperform?  I can submit internal feedback if you provide repro steps.

Some addresses do return a "Bing Map".  We are working on improving the coverage and know that we can make it better.  http://www.bing.com/search?q=2800+W+Big+Beaver+Rd%2C+Troy%2C+MI&form=QBRE

Thanks for the feedback.