Midline formation in insects occurs in a way that is fundamentally different way from that in vertebrates. In Drosophila a dorsal organizing region exists but it is not located on the blastopore and it does not elongate the midline. No ring-to-rod conversion occurs as in vertebrates and no moving prechordal plate exists. The midline has from the beginning the full AP extension.
On the basis of the data available at those times, a very different model has been proposed for insects (Meinhardt, 1989), which found much support by recent observations. A spot-like organizer is assumed that inhibits a stripe-forming system. The stripe forms, therefore, at the opposite ventral site, has the full length but sharpens in the course of time. This is illustrated in the following simulation.
(for equation and parameter click here)
Meanwhile the two main implications of this proposal, the simultaneous midline formation along the AP axis and its refinement along the DV axis, have been most clearly demonstrated in the short germ insect Tribolium (Chen et al.):
While in vertebrates, the midline and the central nervous system forms at the same side as the organizer (dorsal), in Drosophila the dorsal organizer restricts midline formation to the opposite (ventral) side. This may provide a new access for explaining the well-known DV-VD conversion [1,2] distinguishing insects and vertebrates.
In Drosophila midline formation depends on a communication between the oocyte nucleus and the surrounding follicle cells [4-6]. At the crucial stage, the oocyte has the shape of a cone (a in the figure below). The nucleus moves from the posterior pole to a marginal position of the circular anterior side , a process that is necessarily connected with a break of symmetry . The position at which the nucleus comes to rest defines the future dorsal position. A repressing influence of the gurken protein produced at this dorsal organizer restricts the expression of the pipe gene to the ventral follicle cells [7-9]. The resulting pipe expression (blue) is much broader than the final midline. By a complex cascade the plateau-like pipe activation in the follicle cells gives rise to the proper ventral midline in the oocyte [4,5].
(for details click here)
A reduced gurken activity leads to a broader pipe stripe. This, however, leads not to a broader midline but to a split of the midline into two parallel midlines with a de-activation at the ventral-most position  (Fig. c ). This is a clear-cut indication that a genuine pattern-forming process is involved in this process (for details of this mechanism and animated simulation click here ).
Meinhardt, H. (1989). Models for positional signalling with application to the dorsoventral patterning of insects and segregation into different cell types. Development (Supplement 1989), 169-180 [PDF]