Upgrading your "game" - state-of-the-art software research and why you should care Robert Feldt Software Engineering Research Lab, BTH 2007-10-17
onsdag 17 oktober 2007
1
onsdag 17 oktober 2007
2
onsdag 17 oktober 2007
3
onsdag 17 oktober 2007
4
We are also humans!
onsdag 17 oktober 2007
5
Your Use of Research • New/Better/Alternative Ideas (N) • Strengthen/Confirm/Refine Beliefs (S) • Contradict Beliefs (X) • Point to Future/New/Missed Areas (F) • Extend/Move Ideas to Your/Other Area (E) • Attracting personnel onsdag 17 oktober 2007
6
No Fancy Pics!
Intentionally left blank! Mysterious...
At least 20 pages!
Short papers are for wimps! We’re dead serious! onsdag 17 oktober 2007
7
November 2007 • Diff AST’s instead of src for change analysis • Patterns for eliciting usability requirements • Recover arch/design facts if not doc’ed • Reliability models incl. expert judgements onsdag 17 oktober 2007
8
October 2007 issue • Choose web tests from models of uses • Proving in models of parallel software • Predicting faults by src code metrics • Statistical modeling of structure and dev behavior in large OO systems
• Categorize component models onsdag 17 oktober 2007
9
Web app testing Choose tests from models of actual uses
Heuristics for selecting subsets of uses based on lattice Eval on 3 systems in 1k-10k LOC range (OSS + Univ) onsdag 17 oktober 2007
10
UoR: Web app testing • Why useful to you? • • • • • • •
onsdag 17 oktober 2007
N: Reduce number of tests with similar coverage N: System design for auto collection and execution S: Trade-off test suite size vs fault detection capability F: Not autotesting!? Not autotesting web apps!? E: Autocapturing and “massaging” non-system tests? E:Varying/Combining uses for strange situations +: Generating new tests based on the models? 11
September 2007 • • •
onsdag 17 oktober 2007
Best copy-paste detection tools to ease bug fixing Models supporting flexible refactoring tools Systematic testing of model-based code gens
12
August 2007 • • • •
onsdag 17 oktober 2007
Preserve meaning from models to embedded code Tool supporting better distributed, 24/7 QA Auto formal docs from dynamic testing & exec Company stock price implications of security bugs
13
Stock price if sec bug • 147 security bug reports for 18 firms • -0.63% (-860M) market value on disclosure • Lose more in competitive markets • Smaller firms lose more • Releasing patch greatly reduces loss onsdag 17 oktober 2007
14
UoR: Stock price if bug • Why useful to you?
onsdag 17 oktober 2007
• • • •
S: Security defect reports hurt you
•
E: Non-sec defects? Usability defects?
S: Find & release patch as quickly as possible S/F: Learn “Crisis Management” in case of media crisis? F: Know about all press about your firm and product, even if small
15
July 2007 • Annotate UML to visualize design patterns • Reduce test suites via program slicing • Better compression of mobile Java bytecode
onsdag 17 oktober 2007
16
May 2007 • Fault distributions in large sw systems • Classifying runtime exec traces • Software licensing and price guidelines • Unclear if general cost models really work onsdag 17 oktober 2007
17
Fault distributions • Three large, successive telecom projects • •
45-90 modules/proj of 10K-100K LOC/module Mostly C, some Java code
• Small number of modules have most faults! • Fault persistence = many early => many later • Not clear if larger size means more faults • Fault densities similar in similar projects onsdag 17 oktober 2007
18
onsdag 17 oktober 2007
19
UoR: Fault distributions • S: Many faults => Expect even more! • S:You+Process+Context determines fault densities
• X: No clear connection size -> faults • E: Does it hold for smaller/my projects?
What do I need to measure to find out?
onsdag 17 oktober 2007
20
Myths vs Stats • “Fail to use real langs/envs” •
onsdag 17 oktober 2007
• •
60% is specific or applied to lang (Java and/or C) Non-mainstream only in one case
“Uses toy examples”
• • •
13% small or toy examples/data 20% “university” examples/data 67% uses industrial and/or OSS examples/date
21
Myths vs Facts • “Not relevant to my situation” • • •
onsdag 17 oktober 2007
Really so? Can you adapt/try idea in your context? Can you expect it to be exactly your context? Let researchers in to make it relevant to you!
22
So why should you care? • Get new ideas/edge, support decisions, deepen thinking, expand horizons
• We are essentially a FREE resource • You Big => We your research ”consultants” • You Small => We your “outsourced” R&D • New knowledge is fun • We mold your future colleagues •
onsdag 17 oktober 2007
Influence us to make something useful out of them ;) 23
What can you do? • At least skim some journals • Sign up on swell.se • Read RSS feeds with plain-text news • Take part in Workshops & Confs • Contact us for collaboration • We want to do this together with you! onsdag 17 oktober 2007
24
Resources •
• • onsdag 17 oktober 2007
Journals
• • • • •
IEEE Transactions on SE (top) ACM Transactions on SW Eng & Methodology (2nd) IEEE Software (lighter) Automated SE journal (growing) Empirical SE (growing)
Conferences:
http://tinyurl.com/ywqt6j
SE Conf Map:
http://ase.csc.ncsu.edu/semap/ 25
BTH/SERL • SERL = Swedens largest SE research group •
Req Eng, Automated V&V, Empirical
• 1 Professor (top 3 in world), 6 PhDs, 10 PhD students
• BTH = Blekinge Tekniska Högskola • • •
onsdag 17 oktober 2007
Focused on IT & Sustainability Largest # of international students in Sweden Bachelor SE, MSc, Master SE, EuroMaster SE 26
onsdag 17 oktober 2007
27
Some SERL Successes • “Fault slip-through model” @ Ericsson • “Req Abstraction Model” @ ABB & DHR •
onsdag 17 oktober 2007
• •
30% less defects overall Improved communication
“Implementation Proposals” @ ABB
• • •
For distributed/global projects Substantially decrease misunderstandings Support knowledge transfer between sites 28
Questions / Discussion • Do you care about Software / SE research? •
Why / Why not?
• How can we be more useful to you? • What can you contribute with? • How create win-win collaboration? onsdag 17 oktober 2007
state-of-the-art software research and why you should ...
Upgrading your "game". - state-of-the-art software research and why you should care. Robert Feldt ... Predicting faults by src code metrics. ⢠Statistical modeling of structure and dev behavior in large OO systems ... Annotate UML to visualize design patterns. ⢠Reduce test suites via program slicing. ⢠Better compression of ...
Oct 17, 2007 - state-of-the-art software research and why you should care. Robert Feldt .... âReq Abstraction Modelâ @ ABB & DHR. ⢠30% less defects overall.
Permission to make digital or hard copies of all or part of this work for personal or .... the beginning of execution and only writes the data to disk once the.
One of the best things about Arch is that it provides the users with the newest software in a form of ... Ubuntu with no display manager or desktop environment and then install your favorite ... Installation Framework. ... package contains all applic
Try one of the apps below to open or edit this item. pdf-13115\why-should-i-believe-why-should-you-believe-a-wake-up-call-to-the-church-by-terry-read.pdf.
any sport, the NHL counts on fans could be to wear your NHL jerseys showing your allegiance to your favorite team or teams. As soon as the rehab. was finished, a southpaw survived behind. Owing to our games will be won by scoring a measly seven point
Page 1 of 1. why-you-should-opt-for-kitchen-splashbacks-canberra.pdf. why-you-should-opt-for-kitchen-splashbacks-canberra.pdf. Open. Extract. Open with.
Enterprise in New York. ... Tinder, a location-based dating application that gained massive adoption a ... effect simply wouldn't develop. ... FROM TWITTER TO TINDER WHY YOU SHOULD STOP LAUNCHING AT TECH CONFERENCES.pdf.
specialist equipment (e.g wok) and common methods of cooking (e.g stir frying)within that country. Grade. 5. As above: but with an outstanding level of care and ...
Evaluated your design clearly and in detail, considering your client. Grade. 5 ... used â and put into your own words. ... resource (poster, leaflet, book, webpage-.
... of the apps below to open or edit this item. pdf-1882\how-congress-works-and-why-you-should-care-by-lee-h-hamilton-2004-02-26-by-lee-h-hamilton.pdf.
Page 1 of 1. Led stage lights and why you. should opt for them. In any event or function, the stage lighting is very crucial and has a significant. role. You need to take a special care when you are organising an event and. leave no stones unturned t
their understanding of what education is. Teachers must also address the problem .... masters in their own household. So they may be humiliated on the job, but ...
software development process and life-cycle on ... the SIGAda Software Development Standards and Ada .... influence all contracting agency personnel, many.
You have used the leaf bio-mimetic fabric and designed a product clearly linked to the waterproofing feature. You have added labels and colour to explain your design idea. Grade 3. You have used the leaf bio-mimetic fabric and designed a 'creative' p
Page 1 of 3. Top 10 Reasons Why You Should Take the EAP Spot. by Austina De Bonte, longtime EAP & AAP parent. 1. If your kid qualified, they are ready academically. The qualification criteria are high. If your kid passed the bar,. they will do fine.