Task 1 Code clean up and refractoring
  
I had managed to port mysql-awss3 to drizzle when I applied this project. This work involved lots of “copy and paste”, which was more about “make it work”. I need go through all existing code to do some necessary refractoring, fix bugs and adjust code style to drizzle standard.

  Task 2 Pass cloud service information into storage engine
  
Authentication information of cloud service like S3 key id and key value is hard-coded for now. I need to figure out a way how to pass it into storage engine. I find that Zimin has begun to solve the same probem and a blueprint has been registered.

  Task 3 Support Rackspace Cloud File

Cloud cursor now have some private member function to do S3 operations like getting bucket list and putting an object. To support Rackspace, I plan to create a new class as cloud service utility with all necessary virtual methods. Two sub classes for S3 and Rackspace are responsible for implementation for these methods. Cousor has a reference of this unitily which is instansiated according to engine option.

There will be a webinar for new drizzle developer on 15th May. The agenda presented by Jay pipes looks great. Most of my concerned problema will be addressed. Thanks Jay!

Looking forward to meeting you drizzle guys in webinar:)

1) multiple column format support

mysql-awss3 seems only support (key,blob) record, while this engine will support more complex record. In this case, item content stored in cloud should be in some format. I will design and implement a row buffer format for mutiple-column record, with other engines(e.g., CSV engine, Heap engine) as reference.

2) index support

To support index, taking AWS S3 for example, we can use two buckets to store data, one for record (say bucket1) and the other for index (say bucket2).For bucket1, item key stores primary key and item content stores data of all columns. For bucket2, item key stores other keys of a table, and item content stores a set of item keys in bucket1, with each one separated by comma. Keys are in KeyTupleFormat. Since keys in buckets2 may not be unique, addition information should be added upon KeyTupleFormat.
In this way, we can issue two REST requests to get one record data given the index. The first request gets item key of record data by searching index, and the second request gets record data by seaching item key.For full table scan, like mysql-awss3, we can fetch all records in sequence by setting “marker” field of each request header as the key of last record fetched.
Or we can use only one bucket. We distinguish column data and index based on prefix of item key in one bucket. The advantage of two buckets is that we can do a full table scan more quickly. Besides, the implementation for the two features above also will be more low-coupling, so I can implement them one by one more easily.

With respect to code implementation, in this project I will

1. Implement a StorageEngine
2. Implement a Cursor
3. Design data structures for table(and index) information like ArchiveShare and TinaShare
4. Add or Rewrite some uitility functions(e.g., HTTP request and response handlers, XML handlers) if necessary.