For the last few years NAND technology and MLC(multilevel cell, 4 bit cells by analog gate technique) made price very attractive(50% reduction each year) to be used in ultralight laptops. However, while SLC(single level cells) can have 100,000 write, life cycles; MLC can only have 3,000 to 10,000 write life cycles. And boot block may need error code correction. Error code correction may need extra clock cycles of 300 to more than 6,000 buffering cycles, depending on BCH or RS(Reed Solomon) codes; while Hamming codes do not need buffering.

This means file system such as ext2/3 or reiser4fs has to be selected based on less repeated write on the same flash memory bits to lengthen the write life cycles of NaND flash memory cards. As NAND continue to use less line width for lithography each year, the life cycles of write process on MLC will improve, same as SLC life cycles. 4 bit cell however will normally have only 1/4 of the life of a single level cell(single bit)?

Puppy Linux took care of USB flash card by only write at the end of a session to bring everything up to date. Reiser4 puts data on the front then in the back of the free memory, alternating use of free memory spaces. But Microsoft tends to rewrite into any small free space to create fragmentations. So, we have to advise Gates and company to schedule free space usage on NAND MLC flash cards?

Footnote:
When flash memory can be set to analog gate characteristic; sequential cells can have PRML data compression from a library table to read out the codes. This is now set at 50% more data compression per year. The cost increase is minimal, since data compression is software, not dependent on the number of silicon transistors. This policy is similar to hdd(60% increase of capacity per year) pricing for many years so far.