Software Engineering 3156

1. Software Engineering 3156 6-Nov-01 #18: Language I, CVEs, and Databases Phil Gross

2. Administrivia • Upcoming events • Today, 4pm, Interschool (7th floor CEPSR) • Graduate School panel • Tomorrow, 6pm, 415 CEPSR • Me doing C++ again, hopefully better this time • Monday, 11-12:15, Interschool • Phil Wadler: should be a good talk 2

3. Also Monday • 6pm, location TBA • Janak gets his wireless on • All you ever wanted to know about wireless comm • Phones, Ethernet, Bluetooth • Unleashing his gadget-freak side 3

4. Mini-Case Studies • iPod installer • Apple screws up big time • Unix is powerful, but Unix is lame • rm -rf "$2Applications/iTunes.app" 2> /dev/null • IRIX 5.1 • How to make a bad product much worse • At great expense • http://yarchive.net/risks/sgi_irix.html 4

5. Collaborative Virtual Environments • Many people are on-line these days • E-mail works • Chat works, but doesn’t scale too well • People want to see other people • Visual systems are highly optimized for 3D • Get vastly more data • Extra dimensions 5

6. Solutions? • Video conferencing • Aka Brady Bunch Model • Doesn’t scale much past two or three • Virtual environments • Avatars • Usually trivial • Environment • Usually static 6

7. Problem Domains • Web, web, web • http://www.cs.brown.edu/memex/ACMCSHT/51/51.html • Military simulation • All military scales, platoon to division • Mechanized and not • Distributed Software Development • CHIME • Complex back-end sources 7

8. Databases in Nutshell • One of the oldest computer applications • Historically, armies of clerks filed and retrieved • Early computers (i.e. IBM) changed this • Founded on inventions to speed 1890 census • Further developments after intro of Social Security in 1935 (26 million records) • As always with computers, primarily military uses at first 8

9. First Database Systems • 1950s: Idea of a database, as hardware-independent entity • 1960: COBOL established • 1968: IBM’s IMS for Apollo project • 1971: Codasyl (Conference on Data Systems Languages) model 9

10. IMS and Codasyl • IMS is hierarchical • Data has parents, children, siblings • Move through a tree-like structure • Codasyl model is a network model • Think arbitrary pointers • Programmer as navigator 10

11. Relational Model • Ted Codd • "A Relational Model of Data for Large Shared Data Banks," 1970 • Data is independent of storage • Queries are non-procedural and specified against the entire data set • Not an immediately successful idea 11

12. Ah, Politics • IBM had massive investments in IMS • “Strategic” product • Codd was at IBM San Jose • Supposed to be building disk drives, not coming up with new software paradigms • 10 years pass • 1973-9: Ingres at Berkeley • Open source success story! • 1974-9: IBM develops System R and SQL 12

13. Modern Databases • Everybody came from Ingres • Oracle, Sybase, Informix • SQL first commercially released by Oracle • Despite having been developed by IBM • 1980: First relational database from IBM • Ingres lives on: first postgres, then postgresql • Blows out mysql IMHO 13

14. So What’s It All About • Data is stored as rows in tables • Tables have columns • Table employee has columns name, address, id-number, and salary • One row is “’phil’, ‘118th st’, 3456, 120000” • Another row is “’janak’, ‘Long Island’, 5678, 15000” • Set of tables organized into Database 14

15. SQL is Easy: Four Commands • Select, Insert, Update, and Delete • Where clauses can get tricky, though • Select * from employee where salary > 20000 • Insert into employee values(‘Suhit’, ‘Roosevelt Island’, 1010, 60000) • Update employee set salary = 120000 where name = ‘Janak’ • Delete from employee where id=1010 15

16. Theoretically Clean • All operations are on tables • All return values are row sets • Query is actually sent as ascii over the wire • Best strategy to answer is computed by database • Structure of database also stored in “system” tables • Table Tables, table Columns 16

17. Joins • Relate two tables based on common data column • Table with author-ID, author-full-name • Table with paper-title, date, author-ID • Print out paper-title, author-full-name • Normalized: author changes name, only gets changed once • Beware Social Security Numbers! 17

18. Why Is This Important • Software Engineering deals with large-scale systems • Virtually all large-scale systems have a database component • Often the centerpiece of a system • And an organization • Also, DBAs have the highest salaries in the IT field 18

19. Also, Remember IMS • Remind you of anything? • Anything else? • Everything comes back in style • I thought IMS was dead, btw; It’s not • IMS serves 200 million end users, managing over 15 billion Gigabytes of production data and processing over 50 billion transactions every day. • One customer reports 3000 days uptime 19

20. Language I: C • Hello, world #include <stdio.h> main() { printf(“hello, world\n”); } • #include thingy not a C language statement • printf not C language either 20

21. To Compile: • cc hello.c • Produces program… • Called a.out • !!?? • cc –o hello hello.c will produce program called hello 21

22. Simple/Stupid Compilation • Compiler can’t find dependencies • Run-time can’t find dynamic libraries • Well, it can now, but not originally • Import becomes a literal include • Compiles to native code • All files resulting from compilation are physically attached to one another to create executable (static linking) 22

23. Like Java • Or rather, Java (and C++) are wholesale ripoffs of C • While, if-else, for, switch-case (original sin) • Operators, % for mod, ++ and --, prefix and postfix, no exponentiation operator 23

24. Datatypes • Integers, and a little floating point • Signed and unsigned char, int, short, long • No booleans: 0=false, nonzero=true • Different sizes • Not precisely defined! • Characters just aliases for bytes • Machine addresses are a native type • “pointers” 24

25. Pointers • Can get address of any variable • The ‘&’ operator • Can value stored at any address • The ‘*’ operator • Declarations: int *ip; /* ip will “point” to an int */ void *vp; /* vp will “point” to something… */ 25

26. More Pointers int x = 7; ip = &x; printf(“%d\n”, *ip); • Prints 7 *ip = 9 /* assignment */ printf(“%d\n”, x); • Prints 9 26

27. Pointer Arithmetic • If ip == 4, after ip++, ip == 8 • assuming ints are four bytes • This is the meaning of the “type” of a pointer • Doesn’t really have arrays, just pointers int a[7]; a ≡ &a[0] ++a ≡ &a[1] 27

28. C Likes Pointers • No strings, just char arrays terminated with 0 • Char pointers flying all over the place in a C program • usually pointing to strings • Note three meanings of 0 • Falseness • End-of-String char • Null pointer (guaranteed invalid) 28

29. The StringCopy Puzzler void strcpy(char *dst, char *src) { while (*dst++ = *src++) ; } • Assignment returns value • Dereference binds tighter than post-increment • Strings are 0 terminated arrays of chars • 0 ≡ false 29

30. C is Minimal • i.e. nearly nonexistent runtime services • No array/pointer bounds checking • Except for processor halt if you try to dereference 0 • Carefree casting • Reinterpret bits as you please: You’re the boss! • No exception mechanism • Certainly no garbage collection 30

31. Memory management • Everything is primitive by default • If you want a block of bytes, just ask void *malloc(int size) int *myString = (int *)malloc(1024 * sizeof(int)); • myString is all mine, until I free it • Exactly once, please • And if I double-free or don’t free? • I’m F*(&%$!ed 31

32. Pointers Not Null-initialized • char *cp; • Copy a bunch of stuff to cp • Perfectly legal • Totally undefined • Disaster • Need to allocate space to cp 32

33. “Right” way • char *cp = (char *) malloc (SOMESIZE); • Now I can “safely” copy stuff to cp • Why all the quotes? • What if SOMESIZE isn’t big enough… 33

34. EZ Aggregation • struct point { • int x; • int y; • }; • struct point pt; • pt.x = 3; pt.y = 4; 34

35. Struct Meets Pointers struct point *ppt = &pt; (*ppt).x == 3 • Or…. ppt->x == 3 • Wacky “arrow” syntax • And then there are function pointers, but we won’t go there… 35