Additional comments:   
  * The contribution here is that they build malloc like API on tmpfs. Existing implementation requires creating multiple copies in the kernel and the user space with tmpfs. The additional copies are eliminated here.

  * The contribution is incremental as compared to the most related work SPHDE. PSHFS uses malloc instead of the filesystem in SPHDE.
 
  * One limitation is that it cannot store data greater than the SWAP.

  * The authors use the term "persistence" differently. Their definition or its implications is not clear from the paper.

  *  In the evaluation of the locate utility in Section 4.3, PSHFS should be compared to tmpfs as well.