The answer might vary according to your context. However, detecting even SNPs are possible now with NGS technologies, thus insertion/deletion detection is not a big thing. I guess, even simple PCRs can detect insertion/deletions in a known genomic region.
If I guess right what Snijesh ask is the issue regarding the algorithmic detection (computational analysis) of long read deletions or insertions from the next generation sequencing data, not the sequencing technologies.
To provide an information whether the sequenced data has experienced an insertions or deletions requires different computational and algorithmic considerations.
If what I have guessed is right, then his question is looking for the algorithm that can for instance, compare virus species (presume to have a very short read as compared with) human (with a very long read). Sequence alignment between these two distantly related species present a very long insertions or deletions and requires a rigorous computational techniques.
If you using NGS, say for example Illumina data then there is no specific limit on Indel size. On the other hand, the algorithm can definitely tune your results.
Once you align to the reference genome. Based on the supporting reads you can detect from single basepair indels (common indels) to large Indels or chromosome rearrangements as in Cancer.