/
NOTES.txt
66 lines (59 loc) · 2.37 KB
/
NOTES.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
A Haskell EDSL for distributed processing.
Uses a EDSL like xml-streamextract to describe processing pipelines.
Can perform process migration by serializing the state of a process at given points and distributing
the serialized state info a process executable across a graph of instances.
Process migration system should utilize the same framework except without any functionality that
relies on the automatic process migration system.
Use STM for *all* IO:
- tag STM operations with the characteristics of the required IO.
- Associate with the STM operations to an implementation of the required IO interface.
- The implementations of the required IO interface correspond to operations for each level in
the memory hierarchy:
- registers
- cache
- ram
- local storage
- network storage
- global storage
Due to technical limitations only the following hierarchy will be considered:
- ram
- local storage
Then
- ram
- local storage
- network storage
- global storage
Compiler support is likely required to even consider registers and cache under this
abstraction.
Construct the accessors for the data type at runtime from the runtime declared structure?
- high performance of the produced accessors
- The required implementation of the accessor can not be determined until runtime.
- serialization properties
- linear access properties
- random access properties
This is turning into a distributed object system. Which is fine but now I need to revisit what
requirements are going to be satisfied and how.
- data marshalling
- Multiple source and destinations supported:
- streams
- network
- Random-access
- RAM
- file
- object system
- data marshalling for instances
- bind properties to objects
- Supports inter-property binding
- data management
- transactional
- can query object graph for instances satisfying properties
- might be required to implement map/reduce
- map/reduce
I would like these features to be incrementally usable by a client in the following feature sets:
- data serialization
- data deserialization
- object marshalling
- object system
- transactional
- reactive layer implementation
- map/reduce