There are mainly two contrasting approaches when it comes to inter-blockchain connectivity; high-level and low-level. Both are equally important, but approach either the fundamentals or numerics of blockchain conglomeration. To create something like an Internet of Coins, we need to use both approaches concurrently and eclectically. (also featured on De-centralize).
It could be compared to the way Michaelangelo worked to create one of his artful statues. Some of the chipping of the stone had to be done with large chisels and firm blows of the hammer. A high level approach to result in the basic and recognizable shapes of the rock being hewn. Yet simultaneously, he had to work with the utmost caution, scraping out the tiniest details to accentuate the shape of muscles and hair. The results of this combination are an artwork.
High-level connectivity focuses mainly on the controlling aspect of value trading. To make sure that one transaction finishes before sharing the access keys to another initiation for transaction can essentially be done from a high-level protocol. For example, it could be an RPC calling script that controls two daemons for different cryptocurrencies to make an allocated trade between them possible. Low-level connectivity delves down to byte-level transactions to make atomic cross-chain trades possible.
In this case two transactions are meticulously crafted using advanced cryptographic programming, by which an atomic script ensures instantaneous escrow between blockchain constructs. Assembly-like language is used to describe this process. Both approaches contain very useful and some deterring aspects, and thus must both be held up to scrutiny. Combining them in tandem is even more of a challenge, as it requires an eclectic overview and understanding of different technologies, their strengths and weaknesses, and how they may by combined most securely and efficiently.
In another comparison it is creating a thing of architectural beauty, like for example the Obuda Synagogue, St. Nicholas Cathedral Kronstadt, or the Taj Mahal. Any one of these three architectural beauties has taken great skill in prestigious, high-level architecture as well as patiently crafted, low-level ornamental detail. Many high-level solutions are potentially compatible even towards systems that very much differ in their underlying implementation.
However, the high-level approach is difficult to properly secure, and needs a parallel cryptographic network to function safely. It also leaves open security holes in every gap between the high-level construct itself and its controlled daemon subject. Perhaps it is also easier to infuse with malware, as high-level programming is often more easily understood, and agile to work with.
The low-level approach is most robust when applied properly, besides being light weight, and can outlast high-level systemic changes. Even so, it requires very skilled programmers and presents tough cryptographic puzzles to solve. In some cases it can lead to a quest for a holy grail that is most difficult, if not near-impossible to reach in a timely manner. And if a vulnerability is ever found at a later stage, it may be very difficult to find those qualified to repair such a weakness. To make an internet of coins possible, we need both approaches, and we must patiently find the way for us as craftsmen to all build it together.