[External] : Re: InputStream/OutputStream concurrency guarantees

[Viktor Klang]

The creation of said signatures was most likely to reduce copying of 
buffers so copying the buffer would likely be going against the design. 
Presumably the party which passes the array to the read/write is in 
control of said array, so if one wants to defensively copy it before 
passing it to the read/write it is definitely possible to do so on a 
case-by-case basis.

One important difference to something like String is that String is 
immutable, so being able to create a mutable String would subvert the 
guarantees of the type—the same is not the case for 
InputStream/OutputStream.

On 2025-10-30 18:10, Florian Weimer wrote:
> * Viktor Klang:
>
>> Also, since there aren't any consistency effects in place, it is not
>> deterministic which changes made by the writer would be seen by the
>> reader/copier.
> Sure, but if a defensive copy were made, at least the output would be
> internally consistent.  This would be similar to what various
> byte[]-accepting String methods already do.
>
> But it seems to me that the consensus seems to be that
> InputStream/OutputStream do not need to do this, and (for example)
> FileInputStream and FileOutputStream could be changed in ways that make
> multiple byte array reads and writes observable through race conditions.
>
> Thanks,
> Florian
>
-- 
Cheers,
√


Viktor Klang
Software Architect, Java Platform Group
Oracle