impact of installation counts on perceived quality: a case study on debian

19
Impact of Installation Counts on Perceived Quality: A Case Study on Debian Israel Herraiz Emad Shihab Thanh H.D. Nguyen Ahmed E. Hassan israel.herraiz@up m.es

Upload: sailqu

Post on 13-Jan-2017

43 views

Category:

Software


0 download

TRANSCRIPT

Page 1: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Impact of Installation Counts on Perceived Quality:

A Case Study on Debian

Israel Herraiz Emad ShihabThanh H.D. NguyenAhmed E. Hassan

[email protected]

Page 2: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Intelligence throughout the lifetime of a software system

From inception to production

http://sail.cs.queensu.ca

Page 3: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

A Unique View ofSoftware Quality in the Large

Debian has an eco-system of 16,000 packagesThousands of involved developers

~500 reported bugs per month Installation and usage are tracked

Page 4: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

The Development Process of Debian

GCC CVS

Mozilla CVS Firefox Browser

ThunderbirdEmail

Upstream Package

GCC C Compiler

GCC CVSDebian BugDatabase

Page 5: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Tracking Installations

Popularity: total install countsRecent Use (Recent 30 days)

Old Use (Beyond 30 days)

Page 6: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Ultimate Debian Database (UDD)

Source Packages

mypkg libmypkg

mypkg

libmypkg-dev

mypkg-common

Type (Game, Dev, Editor)Priority (Essential , Optional)

Bugs (fixed, reported, forwarded)Reporter (email)

Binary Packages

Page 7: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Sub-linear growth (X1/constant)

Log(installations)

Log(

Bugs

)

Page 8: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

More installations More Bugs

Log(installations)

Log(

Bugs

)

Page 9: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Pattern holds for optional packages

Log(installations)

Log(

Bugs

)

Page 10: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Essential packages

Log(installations)

Log(

Bugs

)

Page 11: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Pattern holds for Fixed Bugs

Log(installations)

Log(

Fixe

d Bu

gs)

Page 12: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Bugs increase with recent use

Recent Use

Log(

fixed

bug

s)

Page 13: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Bug profiles differ between package types

Technical users Casual users

Software Development Games

Page 14: Impact of Installation Counts on Perceived Quality: A Case Study on Debian
Page 15: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

“With many eyes, all bugs become shallow”Eric Raymond

Page 16: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Bug Management OverheadVery low % of bugs are fixed for widely installed packages

Log(installations)

% o

f Fix

ed B

ugs

Page 17: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Sub-linear growth (X1/constant)

Log(installations)

Log(

Bugs

)

Page 18: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Is there a representative set of packages?

Log (installations)

Log(

Bugs

)

Page 19: Impact of Installation Counts on Perceived Quality: A Case Study on Debian

Try this at home http://purl.org/net/who/iht/wcre2011