6.033 Review Session - Naming

• Questions on Naming: pages 10-12 through 10-14 of review packet
• Errors:
  • Question 5.5 - answer is (1996-2-1b) not (1996-2-1a)
  • Question 5.7 - answer is (1987-1-2a) not (1985-2-2)
• Relevant Papers:
  • Class Notes Chapter 5
  • Grapevine
  • NAT

Naming

• reference - used to refer to objects
• used for:
  • sharing
  • interconnection
  • location

In reference to Figure 5.2 in the class notes:

• Name Space - set of possible names.
• Universe of Values - set of all possible values being referenced. Can be objects, or other names.
• Context - additional data needed to resolve a name. (often table of name-> value mappings)
• Name-mapping algorithm - provides actual mapping/association between name space and universe of values.

1. Simple table lookup
   • table of {name, value} pair
   • Ex: telephone book, page table in VM system
2. Path name resolution
   • absolute - uses wired-in context reference to resolve (like root)
   • relative - use different, default context to resolve
   • Ex: web.mit.edu
     /usr/bin/emacs
3. Search
   • Multiple table lookups
   • ex: setting PATH environment variable to specify where programs are to be found.
   • search not perfect, can return "wrong" value

• Important distinction:
  • Context is the actual set of bindings/mappings of names to values
  • Context Reference is the name of the context
• default - wired in or dynamically determined from environment
• explicit - specified for particular object or "packaged" in name (qualified name)

• metadata - data about the data. relevant information about the object, can be contained within the object or outside of it.
• overloading - putting more into name than is needed for reference.
  • ex: .jpg extension to show file type

• protocol for obtaining name
• recursive - need to know 1st place to search to start

• uses several layers, hierarchy, corresponding to physical layers (hardware up to software)
  
  
• blocks on storage device
• inodes - group blocks into files
  
  • inode structure - metadata, then block IDs
  • last two entries of inode point to indirect and doubly indirect blocks instead of data blocks
  • indirect block - points to a block that is just a list of block IDs
  • doubly indirect block - points to a block that lists a bunch of block IDs of indirect blocks.
  • indirect and doubly indirect blocks increase the number of blocks that can be mapped to a file
• directories - organize inodes
  
  • name to inode mappings
  • implemented as file
  • can contain entries for other directories
• went over Question 5.1 on p.10-12. Answers in back of review packet.

• resolution between user-readable names and Internet addresses
• hierarchical - each name server responsible for specific domain/part of name space
• iterative resolution of ginger.lcs.mit.edu
  1. ask root name server
  2. root name server returns "edu is at __"
  3. ask edu name server
  4. edu name server returns "mit.edu is at __"
  5. ask mit.edu name server
  6. mit.edu name server returns "lcs.mit.edu is at __"
  7. ask lcs.mit.edu name server
  8. lcs.mit.edu name server returns "ginger.lcs.mit.edu is at __"
• can also do recursive resolution - ask a name server, it does the rest of the lookups for you and then returns the answer
• if do recursive, can cache responses in that name server so that it can respond to future queries with less delay.
• DNS also has replication for robustness, allows synonyms - all in notes.

Also, make sure to study URL and anything that wasn't covered here. Make sure you have a familiarity with the terminology and are comfortable how it is used and where it applies to various systems. Lots of examples to help you get a feel for this in the notes.